diff --git "a/exp/log/log-train-2023-03-20-17-19-56-1" "b/exp/log/log-train-2023-03-20-17-19-56-1" new file mode 100644--- /dev/null +++ "b/exp/log/log-train-2023-03-20-17-19-56-1" @@ -0,0 +1,25403 @@ +2023-03-20 17:19:56,400 INFO [train.py:971] (1/2) Training started +2023-03-20 17:19:56,400 INFO [train.py:981] (1/2) Device: cuda:1 +2023-03-20 17:19:56,646 INFO [lexicon.py:168] (1/2) Loading pre-compiled data/lang_char/Linv.pt +2023-03-20 17:19:56,675 INFO [train.py:993] (1/2) {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, 'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.22', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '96c9a2aece2a3a7633da07740e24fa3d96f5498c', 'k2-git-date': 'Thu Nov 10 08:14:02 2022', 'lhotse-version': '1.13.0.dev+git.527d964.clean', 'torch-version': '1.12.1', 'torch-cuda-available': True, 'torch-cuda-version': '11.6', 'python-version': '3.8', 'icefall-git-branch': 'aishell_zipformer', 'icefall-git-sha1': '8337628-dirty', 'icefall-git-date': 'Mon Mar 20 16:20:05 2023', 'icefall-path': '/ceph-data4/yangxiaoyu/softwares/icefall_development/icefall_aishell_zipformer', 'k2-path': '/ceph-data4/yangxiaoyu/softwares/anaconda3/envs/k2_latest/lib/python3.8/site-packages/k2/__init__.py', 'lhotse-path': '/ceph-data4/yangxiaoyu/softwares/lhotse_development/lhotse_random_padding_left/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-7-1218101249-5d97868c7c-v8ngc', 'IP address': '10.177.77.18'}, 'world_size': 2, 'master_port': 12354, 'tensorboard': True, 'num_epochs': 50, 'start_epoch': 1, 'start_batch': 0, 'exp_dir': PosixPath('pruned_transducer_stateless7/exp'), 'lang_dir': PosixPath('data/lang_char'), 'base_lr': 0.05, 'lr_batches': 5000, 'lr_epochs': 6, 'context_size': 1, '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': 4000, 'keep_last_k': 30, 'average_period': 200, 'use_fp16': True, 'num_encoder_layers': '2,4,3,2,4', 'feedforward_dims': '1024,1024,2048,2048,1024', 'nhead': '8,8,8,8,8', 'encoder_dims': '384,384,384,384,384', 'attention_dims': '192,192,192,192,192', 'encoder_unmasked_dims': '256,256,256,256,256', 'zipformer_downsampling_factors': '1,2,4,8,2', 'cnn_module_kernels': '31,31,31,31,31', 'decoder_dim': 512, 'joiner_dim': 512, 'max_duration': 300, 'bucketing_sampler': True, 'num_buckets': 30, 'shuffle': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'manifest_dir': PosixPath('data/fbank'), 'on_the_fly_feats': False, 'blank_id': 0, 'vocab_size': 4336} +2023-03-20 17:19:56,675 INFO [train.py:995] (1/2) About to create model +2023-03-20 17:19:57,352 INFO [zipformer.py:178] (1/2) At encoder stack 4, which has downsampling_factor=2, we will combine the outputs of layers 1 and 3, with downsampling_factors=2 and 8. +2023-03-20 17:19:57,455 INFO [train.py:999] (1/2) Number of model parameters: 77741923 +2023-03-20 17:19:59,467 INFO [train.py:1014] (1/2) Using DDP +2023-03-20 17:20:00,492 INFO [aishell.py:39] (1/2) About to get train cuts from data/fbank/aishell_cuts_train.jsonl.gz +2023-03-20 17:20:02,449 INFO [asr_datamodule.py:161] (1/2) Enable MUSAN +2023-03-20 17:20:02,449 INFO [asr_datamodule.py:171] (1/2) Enable SpecAugment +2023-03-20 17:20:02,449 INFO [asr_datamodule.py:172] (1/2) Time warp factor: 80 +2023-03-20 17:20:02,449 INFO [asr_datamodule.py:182] (1/2) Num frame mask: 10 +2023-03-20 17:20:02,449 INFO [asr_datamodule.py:195] (1/2) About to create train dataset +2023-03-20 17:20:02,449 INFO [asr_datamodule.py:223] (1/2) Using DynamicBucketingSampler. +2023-03-20 17:20:02,478 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 17:20:02,526 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 17:20:02,826 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 17:20:02,841 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 17:20:03,004 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 17:20:03,019 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 17:20:03,029 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 17:20:03,141 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 17:20:03,226 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 17:20:03,241 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 17:20:03,258 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 17:20:03,418 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 17:20:03,424 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 17:20:03,431 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 17:20:03,888 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 17:20:03,891 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 17:20:03,948 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 17:20:04,804 INFO [asr_datamodule.py:236] (1/2) About to create train dataloader +2023-03-20 17:20:04,806 INFO [aishell.py:45] (1/2) About to get valid cuts from data/fbank/aishell_cuts_dev.jsonl.gz +2023-03-20 17:20:04,809 INFO [asr_datamodule.py:249] (1/2) About to create dev dataset +2023-03-20 17:20:05,282 INFO [asr_datamodule.py:266] (1/2) About to create dev dataloader +2023-03-20 17:20:05,283 INFO [train.py:1129] (1/2) start training from epoch 1 +2023-03-20 17:20:13,676 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 17:20:13,733 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 17:20:13,759 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 17:20:13,775 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 17:20:13,942 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 17:20:13,957 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 17:20:13,968 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 17:20:14,062 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 17:20:14,139 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 17:20:14,153 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 17:20:14,170 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 17:20:14,772 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 17:20:14,778 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 17:20:14,791 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 17:20:14,928 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 17:20:14,931 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 17:20:14,988 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 17:20:19,860 INFO [train.py:901] (1/2) Epoch 1, batch 0, loss[loss=5.342, simple_loss=8.846, pruned_loss=9.147, over 7287.00 frames. ], tot_loss[loss=5.342, simple_loss=8.846, pruned_loss=9.147, over 7287.00 frames. ], batch size: 66, lr: 2.50e-02, grad_scale: 2.0 +2023-03-20 17:20:19,860 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 17:20:45,999 INFO [train.py:935] (1/2) Epoch 1, validation: loss=4.989, simple_loss=8.202, pruned_loss=8.839, over 1622729.00 frames. +2023-03-20 17:20:46,000 INFO [train.py:936] (1/2) Maximum memory allocated so far is 8207MB +2023-03-20 17:20:47,597 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5.0, num_to_drop=2, layers_to_drop={1, 2} +2023-03-20 17:20:51,641 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 17:20:52,884 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3291, 4.3282, 4.3275, 4.3327, 4.3278, 4.3232, 4.2900, 4.3205], + device='cuda:1'), covar=tensor([0.0010, 0.0009, 0.0011, 0.0011, 0.0010, 0.0013, 0.0028, 0.0021], + device='cuda:1'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], + device='cuda:1'), out_proj_covar=tensor([8.7471e-06, 8.8412e-06, 8.6669e-06, 8.8662e-06, 8.6907e-06, 8.7336e-06, + 8.7569e-06, 8.8802e-06], device='cuda:1') +2023-03-20 17:20:54,308 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=23.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:20:54,354 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9679, 3.9556, 3.9554, 3.9724, 3.9735, 3.9246, 3.9080, 3.8839], + device='cuda:1'), covar=tensor([0.0015, 0.0016, 0.0012, 0.0021, 0.0014, 0.0027, 0.0014, 0.0015], + device='cuda:1'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], + device='cuda:1'), out_proj_covar=tensor([8.6288e-06, 8.7306e-06, 8.7445e-06, 8.6787e-06, 8.9305e-06, 8.7906e-06, + 8.8032e-06, 8.8042e-06], device='cuda:1') +2023-03-20 17:21:00,287 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 17:21:05,507 INFO [train.py:901] (1/2) Epoch 1, batch 50, loss[loss=0.5736, simple_loss=0.8881, pruned_loss=1.054, over 7336.00 frames. ], tot_loss[loss=1.282, simple_loss=2.159, pruned_loss=1.886, over 325939.07 frames. ], batch size: 59, lr: 2.75e-02, grad_scale: 1.0 +2023-03-20 17:21:06,264 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 17:21:07,112 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4535, 4.3894, 4.4081, 4.4383, 4.1383, 4.4482, 4.3279, 4.2736], + device='cuda:1'), covar=tensor([0.0016, 0.0017, 0.0008, 0.0022, 0.0012, 0.0022, 0.0011, 0.0016], + device='cuda:1'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], + device='cuda:1'), out_proj_covar=tensor([8.9534e-06, 9.0605e-06, 9.0850e-06, 9.0300e-06, 9.2537e-06, 9.0977e-06, + 9.0707e-06, 9.1317e-06], device='cuda:1') +2023-03-20 17:21:07,805 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 17:21:09,662 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 17:21:15,432 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=111.84 vs. limit=5.0 +2023-03-20 17:21:17,555 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:21:17,690 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=67.42 vs. limit=5.0 +2023-03-20 17:21:22,003 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.7620, 5.7620, 5.7617, 5.7611, 5.7622, 5.7619, 5.7610, 5.7611], + device='cuda:1'), covar=tensor([4.7468e-04, 8.7819e-05, 5.4602e-04, 4.0912e-04, 6.7338e-04, 8.3863e-04, + 3.4648e-04, 1.9603e-04], device='cuda:1'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], + device='cuda:1'), out_proj_covar=tensor([9.0819e-06, 9.0813e-06, 9.0345e-06, 8.8448e-06, 9.0682e-06, 9.1063e-06, + 8.8696e-06, 8.8928e-06], device='cuda:1') +2023-03-20 17:21:22,906 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=109.41 vs. limit=5.0 +2023-03-20 17:21:23,066 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 17:21:23,436 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 17:21:24,935 INFO [train.py:901] (1/2) Epoch 1, batch 100, loss[loss=0.5121, simple_loss=0.754, pruned_loss=0.929, over 7297.00 frames. ], tot_loss[loss=0.8582, simple_loss=1.392, pruned_loss=1.371, over 575032.63 frames. ], batch size: 86, lr: 3.00e-02, grad_scale: 1.0 +2023-03-20 17:21:25,321 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.002e+01 5.426e+01 1.031e+02 2.944e+02 4.647e+03, threshold=2.062e+02, percent-clipped=0.0 +2023-03-20 17:21:25,569 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=15.90 vs. limit=2.0 +2023-03-20 17:21:29,264 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=4.98 vs. limit=2.0 +2023-03-20 17:21:36,305 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5992, 4.5993, 4.5993, 4.5993, 4.5992, 4.5993, 4.5993, 4.5993], + device='cuda:1'), covar=tensor([1.7025e-04, 1.2989e-04, 1.8338e-04, 1.3773e-04, 1.0568e-04, 1.5025e-04, + 8.7393e-05, 1.3839e-04], device='cuda:1'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], + device='cuda:1'), out_proj_covar=tensor([9.0724e-06, 8.9302e-06, 8.8676e-06, 8.8177e-06, 8.8960e-06, 8.8211e-06, + 8.8772e-06, 8.9660e-06], device='cuda:1') +2023-03-20 17:21:39,370 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0180, 4.4638, 4.4196, 4.4645, 4.0423, 4.3597, 4.3468, 4.2981], + device='cuda:1'), covar=tensor([0.0023, 0.0012, 0.0017, 0.0011, 0.0027, 0.0021, 0.0020, 0.0031], + device='cuda:1'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], + device='cuda:1'), out_proj_covar=tensor([8.9557e-06, 9.0445e-06, 8.8562e-06, 9.1917e-06, 8.9954e-06, 9.0130e-06, + 8.9940e-06, 9.0282e-06], device='cuda:1') +2023-03-20 17:21:41,990 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144.0, num_to_drop=2, layers_to_drop={1, 2} +2023-03-20 17:21:44,523 INFO [train.py:901] (1/2) Epoch 1, batch 150, loss[loss=0.4935, simple_loss=0.696, pruned_loss=0.8665, over 7283.00 frames. ], tot_loss[loss=0.7118, simple_loss=1.115, pruned_loss=1.181, over 767076.72 frames. ], batch size: 66, lr: 3.25e-02, grad_scale: 1.0 +2023-03-20 17:21:51,070 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=75.81 vs. limit=5.0 +2023-03-20 17:21:58,252 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.91 vs. limit=2.0 +2023-03-20 17:21:59,673 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.07 vs. limit=2.0 +2023-03-20 17:22:05,995 INFO [train.py:901] (1/2) Epoch 1, batch 200, loss[loss=0.4935, simple_loss=0.6846, pruned_loss=0.7937, over 7268.00 frames. ], tot_loss[loss=0.6328, simple_loss=0.9621, pruned_loss=1.053, over 917581.21 frames. ], batch size: 66, lr: 3.50e-02, grad_scale: 1.0 +2023-03-20 17:22:06,365 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.567e+01 2.396e+01 3.102e+01 4.606e+01 1.207e+02, threshold=6.203e+01, percent-clipped=0.0 +2023-03-20 17:22:10,600 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 17:22:13,972 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 17:22:14,257 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.17 vs. limit=2.0 +2023-03-20 17:22:18,534 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 17:22:25,478 INFO [train.py:901] (1/2) Epoch 1, batch 250, loss[loss=0.5081, simple_loss=0.6898, pruned_loss=0.7663, over 7327.00 frames. ], tot_loss[loss=0.5865, simple_loss=0.8699, pruned_loss=0.959, over 1032657.38 frames. ], batch size: 54, lr: 3.75e-02, grad_scale: 1.0 +2023-03-20 17:22:27,740 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 17:22:34,587 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=42.50 vs. limit=5.0 +2023-03-20 17:22:38,012 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=3.56 vs. limit=2.0 +2023-03-20 17:22:43,149 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:22:43,812 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 17:22:44,627 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300.0, num_to_drop=2, layers_to_drop={0, 1} +2023-03-20 17:22:44,922 INFO [train.py:901] (1/2) Epoch 1, batch 300, loss[loss=0.4572, simple_loss=0.6048, pruned_loss=0.6575, over 7267.00 frames. ], tot_loss[loss=0.5561, simple_loss=0.8053, pruned_loss=0.8861, over 1124320.73 frames. ], batch size: 47, lr: 4.00e-02, grad_scale: 1.0 +2023-03-20 17:22:45,281 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.093e+01 3.024e+01 4.407e+01 5.736e+01 1.326e+02, threshold=8.814e+01, percent-clipped=19.0 +2023-03-20 17:22:50,411 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 17:23:01,453 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9136, 3.8848, 3.9209, 3.9091, 3.9184, 3.9159, 3.9161, 3.9183], + device='cuda:1'), covar=tensor([0.0038, 0.0034, 0.0043, 0.0042, 0.0046, 0.0033, 0.0037, 0.0045], + device='cuda:1'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], + device='cuda:1'), out_proj_covar=tensor([8.8360e-06, 8.7275e-06, 8.9754e-06, 8.7995e-06, 9.0636e-06, 8.7374e-06, + 9.0366e-06, 8.8369e-06], device='cuda:1') +2023-03-20 17:23:03,914 INFO [train.py:901] (1/2) Epoch 1, batch 350, loss[loss=0.4957, simple_loss=0.6384, pruned_loss=0.6843, over 7315.00 frames. ], tot_loss[loss=0.5384, simple_loss=0.761, pruned_loss=0.8331, over 1195222.00 frames. ], batch size: 49, lr: 4.25e-02, grad_scale: 1.0 +2023-03-20 17:23:06,253 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=357.0, num_to_drop=2, layers_to_drop={0, 1} +2023-03-20 17:23:12,560 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=8.97 vs. limit=5.0 +2023-03-20 17:23:12,803 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4035, 4.3742, 4.3854, 4.4159, 4.4078, 4.4164, 4.3841, 4.1105], + device='cuda:1'), covar=tensor([0.0055, 0.0089, 0.0085, 0.0050, 0.0055, 0.0068, 0.0089, 0.0359], + device='cuda:1'), in_proj_covar=tensor([0.0010, 0.0010, 0.0009, 0.0010, 0.0009, 0.0010, 0.0009, 0.0010], + device='cuda:1'), out_proj_covar=tensor([9.0873e-06, 9.4157e-06, 9.1786e-06, 9.5333e-06, 9.4374e-06, 9.5158e-06, + 9.2763e-06, 9.4016e-06], device='cuda:1') +2023-03-20 17:23:16,946 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 17:23:18,057 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=387.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:23:23,744 INFO [train.py:901] (1/2) Epoch 1, batch 400, loss[loss=0.5282, simple_loss=0.6652, pruned_loss=0.6975, over 7288.00 frames. ], tot_loss[loss=0.5289, simple_loss=0.7294, pruned_loss=0.7929, over 1251062.70 frames. ], batch size: 70, lr: 4.50e-02, grad_scale: 2.0 +2023-03-20 17:23:24,106 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.730e+01 3.441e+01 4.515e+01 6.284e+01 2.215e+02, threshold=9.029e+01, percent-clipped=10.0 +2023-03-20 17:23:38,542 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=439.0, num_to_drop=2, layers_to_drop={1, 2} +2023-03-20 17:23:41,872 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=448.0, num_to_drop=2, layers_to_drop={0, 1} +2023-03-20 17:23:42,913 INFO [train.py:901] (1/2) Epoch 1, batch 450, loss[loss=0.5117, simple_loss=0.6257, pruned_loss=0.6563, over 7299.00 frames. ], tot_loss[loss=0.5238, simple_loss=0.705, pruned_loss=0.76, over 1291956.94 frames. ], batch size: 80, lr: 4.75e-02, grad_scale: 2.0 +2023-03-20 17:23:47,973 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 17:23:48,325 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 17:23:49,196 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3743, 3.4374, 3.4203, 3.4495, 3.4472, 3.4185, 3.4451, 3.4089], + device='cuda:1'), covar=tensor([0.0164, 0.0109, 0.0105, 0.0095, 0.0094, 0.0102, 0.0119, 0.0112], + device='cuda:1'), in_proj_covar=tensor([0.0010, 0.0009, 0.0009, 0.0009, 0.0010, 0.0009, 0.0010, 0.0010], + device='cuda:1'), out_proj_covar=tensor([9.1937e-06, 8.8278e-06, 9.0386e-06, 9.0645e-06, 9.0843e-06, 8.8810e-06, + 9.0991e-06, 9.0420e-06], device='cuda:1') +2023-03-20 17:24:01,952 INFO [train.py:901] (1/2) Epoch 1, batch 500, loss[loss=0.4899, simple_loss=0.595, pruned_loss=0.5911, over 7355.00 frames. ], tot_loss[loss=0.5225, simple_loss=0.6873, pruned_loss=0.7318, over 1326832.64 frames. ], batch size: 73, lr: 4.99e-02, grad_scale: 2.0 +2023-03-20 17:24:02,311 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.353e+01 4.338e+01 5.607e+01 7.790e+01 1.498e+02, threshold=1.121e+02, percent-clipped=18.0 +2023-03-20 17:24:12,193 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 17:24:13,331 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 17:24:13,678 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 17:24:15,137 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 17:24:18,443 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 17:24:21,053 INFO [train.py:901] (1/2) Epoch 1, batch 550, loss[loss=0.4992, simple_loss=0.5974, pruned_loss=0.5763, over 7279.00 frames. ], tot_loss[loss=0.5203, simple_loss=0.6712, pruned_loss=0.7014, over 1352676.03 frames. ], batch size: 70, lr: 4.98e-02, grad_scale: 2.0 +2023-03-20 17:24:25,654 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=562.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:24:27,742 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 17:24:27,833 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=568.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:24:33,242 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 17:24:35,817 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 17:24:35,906 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=590.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:24:40,196 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=600.0, num_to_drop=2, layers_to_drop={0, 2} +2023-03-20 17:24:40,468 INFO [train.py:901] (1/2) Epoch 1, batch 600, loss[loss=0.4364, simple_loss=0.5262, pruned_loss=0.4677, over 7175.00 frames. ], tot_loss[loss=0.5128, simple_loss=0.6521, pruned_loss=0.6634, over 1372784.07 frames. ], batch size: 39, lr: 4.98e-02, grad_scale: 2.0 +2023-03-20 17:24:40,840 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 5.139e+01 1.040e+02 1.577e+02 2.315e+02 8.829e+02, threshold=3.155e+02, percent-clipped=69.0 +2023-03-20 17:24:41,635 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 17:24:48,020 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=7.17 vs. limit=5.0 +2023-03-20 17:24:49,328 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=623.0, num_to_drop=2, layers_to_drop={0, 1} +2023-03-20 17:24:51,619 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=629.0, num_to_drop=2, layers_to_drop={1, 2} +2023-03-20 17:24:53,801 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 17:24:59,236 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=648.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:25:00,341 INFO [train.py:901] (1/2) Epoch 1, batch 650, loss[loss=0.4176, simple_loss=0.5071, pruned_loss=0.4174, over 7193.00 frames. ], tot_loss[loss=0.5023, simple_loss=0.632, pruned_loss=0.6219, over 1388558.20 frames. ], batch size: 39, lr: 4.98e-02, grad_scale: 2.0 +2023-03-20 17:25:00,455 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=651.0, num_to_drop=2, layers_to_drop={1, 2} +2023-03-20 17:25:00,717 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 17:25:00,771 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=652.0, num_to_drop=2, layers_to_drop={1, 3} +2023-03-20 17:25:12,940 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 17:25:14,250 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=3.28 vs. limit=2.0 +2023-03-20 17:25:18,781 INFO [train.py:901] (1/2) Epoch 1, batch 700, loss[loss=0.3841, simple_loss=0.4708, pruned_loss=0.358, over 7333.00 frames. ], tot_loss[loss=0.4884, simple_loss=0.6096, pruned_loss=0.5781, over 1402164.48 frames. ], batch size: 44, lr: 4.98e-02, grad_scale: 2.0 +2023-03-20 17:25:19,157 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.970e+02 2.813e+02 3.918e+02 8.895e+02, threshold=5.626e+02, percent-clipped=42.0 +2023-03-20 17:25:19,209 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 17:25:27,708 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.83 vs. limit=2.0 +2023-03-20 17:25:33,327 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=739.0, num_to_drop=2, layers_to_drop={0, 2} +2023-03-20 17:25:35,286 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=743.0, num_to_drop=2, layers_to_drop={1, 3} +2023-03-20 17:25:37,994 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 17:25:38,688 INFO [train.py:901] (1/2) Epoch 1, batch 750, loss[loss=0.3951, simple_loss=0.4729, pruned_loss=0.3623, over 7340.00 frames. ], tot_loss[loss=0.475, simple_loss=0.5888, pruned_loss=0.5375, over 1410903.56 frames. ], batch size: 44, lr: 4.97e-02, grad_scale: 2.0 +2023-03-20 17:25:38,707 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 17:25:48,265 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 17:25:51,929 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 17:25:51,992 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=787.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:25:56,024 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 17:25:58,329 INFO [train.py:901] (1/2) Epoch 1, batch 800, loss[loss=0.4218, simple_loss=0.506, pruned_loss=0.3666, over 7219.00 frames. ], tot_loss[loss=0.4617, simple_loss=0.5686, pruned_loss=0.4998, over 1416807.69 frames. ], batch size: 50, lr: 4.97e-02, grad_scale: 4.0 +2023-03-20 17:25:58,337 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 17:25:58,690 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.725e+02 3.452e+02 4.919e+02 8.420e+02, threshold=6.904e+02, percent-clipped=17.0 +2023-03-20 17:26:06,084 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 17:26:16,075 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=847.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:26:17,482 INFO [train.py:901] (1/2) Epoch 1, batch 850, loss[loss=0.3734, simple_loss=0.453, pruned_loss=0.3042, over 7301.00 frames. ], tot_loss[loss=0.4471, simple_loss=0.548, pruned_loss=0.4631, over 1419949.75 frames. ], batch size: 83, lr: 4.96e-02, grad_scale: 4.0 +2023-03-20 17:26:18,072 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-20 17:26:20,154 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 17:26:20,525 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 17:26:24,600 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 17:26:27,207 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 17:26:36,320 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6760, 3.3137, 3.3376, 3.1389, 3.3678, 3.1877, 3.3766, 2.8391], + device='cuda:1'), covar=tensor([0.2719, 0.4371, 0.3778, 0.3936, 0.4591, 0.4464, 0.3610, 0.5843], + device='cuda:1'), in_proj_covar=tensor([0.0026, 0.0027, 0.0027, 0.0028, 0.0027, 0.0028, 0.0026, 0.0030], + device='cuda:1'), out_proj_covar=tensor([2.2435e-05, 2.5128e-05, 2.3750e-05, 2.5946e-05, 2.4790e-05, 2.5300e-05, + 2.3712e-05, 2.8853e-05], device='cuda:1') +2023-03-20 17:26:36,617 INFO [train.py:901] (1/2) Epoch 1, batch 900, loss[loss=0.36, simple_loss=0.4345, pruned_loss=0.2824, over 7351.00 frames. ], tot_loss[loss=0.4355, simple_loss=0.5311, pruned_loss=0.4312, over 1426687.35 frames. ], batch size: 44, lr: 4.96e-02, grad_scale: 4.0 +2023-03-20 17:26:36,974 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 2.661e+02 3.954e+02 5.359e+02 9.011e+02, threshold=7.908e+02, percent-clipped=11.0 +2023-03-20 17:26:39,302 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=908.0, num_to_drop=2, layers_to_drop={0, 2} +2023-03-20 17:26:43,027 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=918.0, num_to_drop=2, layers_to_drop={0, 3} +2023-03-20 17:26:45,086 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.15 vs. limit=2.0 +2023-03-20 17:26:45,208 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=924.0, num_to_drop=2, layers_to_drop={0, 3} +2023-03-20 17:26:53,861 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=946.0, num_to_drop=2, layers_to_drop={0, 1} +2023-03-20 17:26:55,311 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 17:26:55,653 INFO [train.py:901] (1/2) Epoch 1, batch 950, loss[loss=0.3863, simple_loss=0.4621, pruned_loss=0.294, over 7335.00 frames. ], tot_loss[loss=0.4274, simple_loss=0.5179, pruned_loss=0.4053, over 1428129.45 frames. ], batch size: 61, lr: 4.96e-02, grad_scale: 4.0 +2023-03-20 17:26:56,101 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=952.0, num_to_drop=2, layers_to_drop={0, 1} +2023-03-20 17:27:06,141 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2599, 3.9627, 3.9531, 4.0635, 4.1011, 3.8662, 3.8007, 3.9649], + device='cuda:1'), covar=tensor([0.1097, 0.1484, 0.1518, 0.0925, 0.0901, 0.1612, 0.1535, 0.0972], + device='cuda:1'), in_proj_covar=tensor([0.0013, 0.0014, 0.0014, 0.0012, 0.0012, 0.0013, 0.0012, 0.0011], + device='cuda:1'), out_proj_covar=tensor([1.1142e-05, 1.1182e-05, 1.1288e-05, 9.7823e-06, 9.1652e-06, 1.0536e-05, + 9.3557e-06, 9.0132e-06], device='cuda:1') +2023-03-20 17:27:12,616 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 17:27:15,033 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=1000.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:27:15,373 INFO [train.py:901] (1/2) Epoch 1, batch 1000, loss[loss=0.4235, simple_loss=0.4829, pruned_loss=0.3307, over 7219.00 frames. ], tot_loss[loss=0.4186, simple_loss=0.5048, pruned_loss=0.3802, over 1433784.68 frames. ], batch size: 93, lr: 4.95e-02, grad_scale: 4.0 +2023-03-20 17:27:15,714 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 3.326e+02 4.317e+02 5.927e+02 1.638e+03, threshold=8.634e+02, percent-clipped=6.0 +2023-03-20 17:27:28,662 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 17:27:30,101 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-03-20 17:27:32,074 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1043.0, num_to_drop=2, layers_to_drop={1, 2} +2023-03-20 17:27:35,075 INFO [train.py:901] (1/2) Epoch 1, batch 1050, loss[loss=0.3477, simple_loss=0.4167, pruned_loss=0.2432, over 7246.00 frames. ], tot_loss[loss=0.4105, simple_loss=0.4925, pruned_loss=0.358, over 1435289.78 frames. ], batch size: 45, lr: 4.95e-02, grad_scale: 4.0 +2023-03-20 17:27:38,360 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 17:27:43,885 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 17:27:47,591 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 17:27:51,122 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=1091.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:27:55,029 INFO [train.py:901] (1/2) Epoch 1, batch 1100, loss[loss=0.3804, simple_loss=0.4487, pruned_loss=0.262, over 7337.00 frames. ], tot_loss[loss=0.4042, simple_loss=0.4827, pruned_loss=0.3382, over 1439762.81 frames. ], batch size: 63, lr: 4.94e-02, grad_scale: 4.0 +2023-03-20 17:27:55,398 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 3.735e+02 4.896e+02 6.513e+02 1.819e+03, threshold=9.792e+02, percent-clipped=10.0 +2023-03-20 17:28:09,548 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 17:28:09,889 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 17:28:14,559 INFO [train.py:901] (1/2) Epoch 1, batch 1150, loss[loss=0.3066, simple_loss=0.349, pruned_loss=0.2137, over 5911.00 frames. ], tot_loss[loss=0.3976, simple_loss=0.4722, pruned_loss=0.3205, over 1440133.23 frames. ], batch size: 25, lr: 4.94e-02, grad_scale: 4.0 +2023-03-20 17:28:15,440 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1153.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:28:19,125 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 17:28:19,861 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 17:28:34,655 INFO [train.py:901] (1/2) Epoch 1, batch 1200, loss[loss=0.3779, simple_loss=0.4423, pruned_loss=0.2447, over 7306.00 frames. ], tot_loss[loss=0.3906, simple_loss=0.4617, pruned_loss=0.3035, over 1438561.74 frames. ], batch size: 86, lr: 4.93e-02, grad_scale: 8.0 +2023-03-20 17:28:35,019 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.042e+02 3.708e+02 4.369e+02 5.950e+02 1.977e+03, threshold=8.739e+02, percent-clipped=4.0 +2023-03-20 17:28:35,472 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1203.0, num_to_drop=2, layers_to_drop={1, 2} +2023-03-20 17:28:40,236 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1214.0, num_to_drop=2, layers_to_drop={2, 3} +2023-03-20 17:28:41,754 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1218.0, num_to_drop=2, layers_to_drop={1, 2} +2023-03-20 17:28:43,959 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1224.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:28:45,733 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 17:28:52,364 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1246.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 17:28:54,252 INFO [train.py:901] (1/2) Epoch 1, batch 1250, loss[loss=0.3542, simple_loss=0.4043, pruned_loss=0.2293, over 7155.00 frames. ], tot_loss[loss=0.3871, simple_loss=0.4547, pruned_loss=0.2904, over 1441505.08 frames. ], batch size: 41, lr: 4.92e-02, grad_scale: 8.0 +2023-03-20 17:28:59,906 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=1266.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:29:02,717 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=1272.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:29:04,299 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 17:29:07,509 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 17:29:08,267 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 17:29:11,473 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=1294.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:29:14,203 INFO [train.py:901] (1/2) Epoch 1, batch 1300, loss[loss=0.3438, simple_loss=0.4003, pruned_loss=0.2095, over 7303.00 frames. ], tot_loss[loss=0.384, simple_loss=0.4482, pruned_loss=0.2789, over 1439162.78 frames. ], batch size: 83, lr: 4.92e-02, grad_scale: 8.0 +2023-03-20 17:29:14,565 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.819e+02 3.657e+02 4.833e+02 5.566e+02 9.665e+02, threshold=9.665e+02, percent-clipped=1.0 +2023-03-20 17:29:18,547 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.24 vs. limit=2.0 +2023-03-20 17:29:18,904 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=6.24 vs. limit=5.0 +2023-03-20 17:29:25,630 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 17:29:27,271 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 17:29:30,074 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 17:29:34,436 INFO [train.py:901] (1/2) Epoch 1, batch 1350, loss[loss=0.3643, simple_loss=0.4236, pruned_loss=0.2153, over 7247.00 frames. ], tot_loss[loss=0.381, simple_loss=0.4419, pruned_loss=0.2681, over 1439145.07 frames. ], batch size: 55, lr: 4.91e-02, grad_scale: 8.0 +2023-03-20 17:29:38,739 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 17:29:55,967 INFO [train.py:901] (1/2) Epoch 1, batch 1400, loss[loss=0.3304, simple_loss=0.3563, pruned_loss=0.2085, over 5823.00 frames. ], tot_loss[loss=0.3774, simple_loss=0.4349, pruned_loss=0.2578, over 1438386.92 frames. ], batch size: 25, lr: 4.91e-02, grad_scale: 8.0 +2023-03-20 17:29:56,338 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.053e+02 4.073e+02 4.817e+02 5.677e+02 9.535e+02, threshold=9.634e+02, percent-clipped=0.0 +2023-03-20 17:30:01,736 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.14 vs. limit=2.0 +2023-03-20 17:30:06,460 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 17:30:09,420 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1434.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:30:13,070 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 +2023-03-20 17:30:13,317 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1439, 3.0481, 3.0257, 2.8650, 2.6959, 3.4887, 3.3291, 3.2290], + device='cuda:1'), covar=tensor([0.1709, 0.1648, 0.1555, 0.2431, 0.1736, 0.1309, 0.1782, 0.1234], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0025, 0.0030, 0.0031, 0.0034, 0.0026, 0.0029, 0.0027], + device='cuda:1'), out_proj_covar=tensor([2.3642e-05, 2.1478e-05, 2.7497e-05, 2.8303e-05, 2.9891e-05, 2.3489e-05, + 2.7984e-05, 2.3100e-05], device='cuda:1') +2023-03-20 17:30:16,030 INFO [train.py:901] (1/2) Epoch 1, batch 1450, loss[loss=0.3769, simple_loss=0.3938, pruned_loss=0.2391, over 6973.00 frames. ], tot_loss[loss=0.376, simple_loss=0.4299, pruned_loss=0.2499, over 1439147.19 frames. ], batch size: 35, lr: 4.90e-02, grad_scale: 8.0 +2023-03-20 17:30:25,505 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 17:30:35,450 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1495.0, num_to_drop=2, layers_to_drop={1, 2} +2023-03-20 17:30:37,664 INFO [train.py:901] (1/2) Epoch 1, batch 1500, loss[loss=0.3977, simple_loss=0.4279, pruned_loss=0.2369, over 7327.00 frames. ], tot_loss[loss=0.3757, simple_loss=0.4262, pruned_loss=0.2432, over 1439772.00 frames. ], batch size: 75, lr: 4.89e-02, grad_scale: 8.0 +2023-03-20 17:30:38,103 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.828e+02 4.191e+02 5.686e+02 6.667e+02 2.069e+03, threshold=1.137e+03, percent-clipped=7.0 +2023-03-20 17:30:38,638 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1503.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:30:39,399 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 17:30:41,146 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1509.0, num_to_drop=2, layers_to_drop={2, 3} +2023-03-20 17:30:41,758 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-03-20 17:30:55,373 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-03-20 17:30:58,655 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 17:30:59,043 INFO [train.py:901] (1/2) Epoch 1, batch 1550, loss[loss=0.3211, simple_loss=0.3462, pruned_loss=0.1855, over 7046.00 frames. ], tot_loss[loss=0.3739, simple_loss=0.4211, pruned_loss=0.236, over 1439587.85 frames. ], batch size: 35, lr: 4.89e-02, grad_scale: 8.0 +2023-03-20 17:30:59,103 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=1551.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:31:05,369 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1566.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:31:20,766 INFO [train.py:901] (1/2) Epoch 1, batch 1600, loss[loss=0.4191, simple_loss=0.4396, pruned_loss=0.2429, over 7350.00 frames. ], tot_loss[loss=0.3745, simple_loss=0.4186, pruned_loss=0.2306, over 1441260.18 frames. ], batch size: 73, lr: 4.88e-02, grad_scale: 8.0 +2023-03-20 17:31:21,139 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.008e+02 4.402e+02 5.060e+02 6.395e+02 1.208e+03, threshold=1.012e+03, percent-clipped=1.0 +2023-03-20 17:31:24,906 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 17:31:25,008 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1611.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:31:25,302 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 17:31:25,350 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4989, 4.1988, 4.2076, 4.3472, 4.4856, 4.3498, 4.3722, 3.9462], + device='cuda:1'), covar=tensor([0.0403, 0.0534, 0.0592, 0.0571, 0.0406, 0.0424, 0.0373, 0.0479], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0047, 0.0057, 0.0040, 0.0050, 0.0047, 0.0039, 0.0049], + device='cuda:1'), out_proj_covar=tensor([3.8165e-05, 4.1311e-05, 5.6723e-05, 3.7817e-05, 4.8083e-05, 4.2044e-05, + 3.6744e-05, 4.5468e-05], device='cuda:1') +2023-03-20 17:31:28,046 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 17:31:31,332 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1627.0, num_to_drop=2, layers_to_drop={0, 1} +2023-03-20 17:31:36,408 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 17:31:39,317 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 17:31:41,599 INFO [train.py:901] (1/2) Epoch 1, batch 1650, loss[loss=0.4672, simple_loss=0.4666, pruned_loss=0.2775, over 6689.00 frames. ], tot_loss[loss=0.3748, simple_loss=0.415, pruned_loss=0.2261, over 1440185.30 frames. ], batch size: 106, lr: 4.87e-02, grad_scale: 8.0 +2023-03-20 17:31:44,290 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9019, 3.6825, 3.8720, 3.9180, 3.9349, 3.6040, 3.9922, 3.7356], + device='cuda:1'), covar=tensor([0.0337, 0.0523, 0.0418, 0.0261, 0.0221, 0.0497, 0.0303, 0.0303], + device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0021, 0.0024, 0.0021, 0.0020, 0.0023, 0.0024, 0.0021], + device='cuda:1'), out_proj_covar=tensor([1.6692e-05, 1.7684e-05, 2.1700e-05, 1.6449e-05, 1.5801e-05, 1.9366e-05, + 2.0086e-05, 1.8028e-05], device='cuda:1') +2023-03-20 17:31:45,545 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2863, 4.2362, 4.1780, 3.9996, 4.1800, 4.2479, 4.0361, 4.3879], + device='cuda:1'), covar=tensor([0.0370, 0.0354, 0.0511, 0.0498, 0.0437, 0.0351, 0.0517, 0.0250], + device='cuda:1'), in_proj_covar=tensor([0.0018, 0.0023, 0.0023, 0.0021, 0.0023, 0.0022, 0.0024, 0.0019], + device='cuda:1'), out_proj_covar=tensor([1.2692e-05, 1.7280e-05, 1.8427e-05, 1.5510e-05, 1.8369e-05, 1.6166e-05, + 1.7668e-05, 1.4202e-05], device='cuda:1') +2023-03-20 17:31:46,661 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 17:31:50,558 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1672.0, num_to_drop=2, layers_to_drop={2, 3} +2023-03-20 17:31:57,212 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 +2023-03-20 17:32:01,071 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 17:32:03,117 INFO [train.py:901] (1/2) Epoch 1, batch 1700, loss[loss=0.3855, simple_loss=0.407, pruned_loss=0.2103, over 7303.00 frames. ], tot_loss[loss=0.3724, simple_loss=0.4103, pruned_loss=0.2193, over 1440186.50 frames. ], batch size: 83, lr: 4.86e-02, grad_scale: 8.0 +2023-03-20 17:32:03,547 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.196e+02 4.136e+02 5.429e+02 6.545e+02 1.463e+03, threshold=1.086e+03, percent-clipped=5.0 +2023-03-20 17:32:04,767 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 17:32:04,852 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6254, 4.3103, 4.2171, 3.9086, 4.0544, 3.6746, 4.1998, 4.2489], + device='cuda:1'), covar=tensor([0.0607, 0.0259, 0.0385, 0.0443, 0.0668, 0.0732, 0.0731, 0.0374], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0045, 0.0045, 0.0057, 0.0054, 0.0049, 0.0046, 0.0048], + device='cuda:1'), out_proj_covar=tensor([3.6495e-05, 3.6881e-05, 3.9084e-05, 5.3017e-05, 4.5334e-05, 4.3558e-05, + 4.0290e-05, 4.5186e-05], device='cuda:1') +2023-03-20 17:32:13,607 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 17:32:24,236 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.14 vs. limit=2.0 +2023-03-20 17:32:24,813 INFO [train.py:901] (1/2) Epoch 1, batch 1750, loss[loss=0.4001, simple_loss=0.4212, pruned_loss=0.2134, over 7360.00 frames. ], tot_loss[loss=0.3723, simple_loss=0.4074, pruned_loss=0.2145, over 1440884.60 frames. ], batch size: 73, lr: 4.86e-02, grad_scale: 8.0 +2023-03-20 17:32:33,815 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 17:32:34,680 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 17:32:40,669 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-03-20 17:32:41,801 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1790.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 17:32:46,791 INFO [train.py:901] (1/2) Epoch 1, batch 1800, loss[loss=0.3556, simple_loss=0.3919, pruned_loss=0.1754, over 7348.00 frames. ], tot_loss[loss=0.3733, simple_loss=0.4055, pruned_loss=0.2106, over 1442069.83 frames. ], batch size: 54, lr: 4.85e-02, grad_scale: 8.0 +2023-03-20 17:32:47,180 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.610e+02 3.716e+02 4.744e+02 6.070e+02 1.750e+03, threshold=9.488e+02, percent-clipped=2.0 +2023-03-20 17:32:50,265 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1809.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:32:53,912 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 17:33:03,017 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1838.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:33:06,300 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 17:33:08,390 INFO [train.py:901] (1/2) Epoch 1, batch 1850, loss[loss=0.2683, simple_loss=0.2949, pruned_loss=0.1296, over 7010.00 frames. ], tot_loss[loss=0.3725, simple_loss=0.4016, pruned_loss=0.2061, over 1440972.55 frames. ], batch size: 35, lr: 4.84e-02, grad_scale: 8.0 +2023-03-20 17:33:11,057 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=1857.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:33:11,105 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1857.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:33:14,516 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 17:33:16,751 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1870.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:33:28,502 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.10 vs. limit=2.0 +2023-03-20 17:33:29,453 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 17:33:29,587 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1899.0, num_to_drop=2, layers_to_drop={0, 2} +2023-03-20 17:33:30,369 INFO [train.py:901] (1/2) Epoch 1, batch 1900, loss[loss=0.3947, simple_loss=0.4078, pruned_loss=0.1997, over 7226.00 frames. ], tot_loss[loss=0.3719, simple_loss=0.3984, pruned_loss=0.2018, over 1440674.91 frames. ], batch size: 93, lr: 4.83e-02, grad_scale: 8.0 +2023-03-20 17:33:30,783 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.126e+02 4.333e+02 5.813e+02 7.332e+02 1.551e+03, threshold=1.163e+03, percent-clipped=10.0 +2023-03-20 17:33:31,961 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.15 vs. limit=2.0 +2023-03-20 17:33:37,832 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1918.0, num_to_drop=2, layers_to_drop={0, 2} +2023-03-20 17:33:39,482 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1922.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 17:33:44,021 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1931.0, num_to_drop=2, layers_to_drop={1, 3} +2023-03-20 17:33:50,293 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 17:33:52,351 INFO [train.py:901] (1/2) Epoch 1, batch 1950, loss[loss=0.3924, simple_loss=0.4052, pruned_loss=0.1941, over 7252.00 frames. ], tot_loss[loss=0.3733, simple_loss=0.397, pruned_loss=0.1987, over 1443885.08 frames. ], batch size: 89, lr: 4.83e-02, grad_scale: 8.0 +2023-03-20 17:33:52,462 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2318, 4.2616, 4.1491, 4.2023, 3.7869, 4.1212, 4.0544, 4.1329], + device='cuda:1'), covar=tensor([0.0208, 0.0121, 0.0152, 0.0128, 0.0401, 0.0238, 0.0227, 0.0177], + device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0031, 0.0032, 0.0033, 0.0037, 0.0033, 0.0033, 0.0033], + device='cuda:1'), out_proj_covar=tensor([2.3433e-05, 2.1749e-05, 2.3150e-05, 2.3319e-05, 3.0533e-05, 2.5033e-05, + 2.2913e-05, 2.3073e-05], device='cuda:1') +2023-03-20 17:33:59,798 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1967.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:34:01,003 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 17:34:04,643 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 17:34:05,038 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 17:34:14,557 INFO [train.py:901] (1/2) Epoch 1, batch 2000, loss[loss=0.3704, simple_loss=0.3907, pruned_loss=0.175, over 7290.00 frames. ], tot_loss[loss=0.3729, simple_loss=0.3944, pruned_loss=0.1947, over 1444672.62 frames. ], batch size: 66, lr: 4.82e-02, grad_scale: 8.0 +2023-03-20 17:34:15,035 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.389e+02 4.542e+02 5.662e+02 7.090e+02 1.677e+03, threshold=1.132e+03, percent-clipped=2.0 +2023-03-20 17:34:20,173 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 17:34:27,770 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4718, 3.6145, 3.5935, 3.9110, 3.5946, 3.4397, 3.5647, 3.7769], + device='cuda:1'), covar=tensor([0.0493, 0.0195, 0.0164, 0.0094, 0.0247, 0.0506, 0.0230, 0.0129], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0033, 0.0030, 0.0028, 0.0030, 0.0029, 0.0031, 0.0029], + device='cuda:1'), out_proj_covar=tensor([2.7548e-05, 2.6666e-05, 2.3523e-05, 2.1567e-05, 2.3268e-05, 2.4214e-05, + 2.3880e-05, 2.2532e-05], device='cuda:1') +2023-03-20 17:34:30,863 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 17:34:32,873 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3140, 3.6283, 3.4143, 3.5729, 3.4715, 3.2003, 3.8233, 3.8774], + device='cuda:1'), covar=tensor([0.0323, 0.0209, 0.0392, 0.0168, 0.0115, 0.0350, 0.0067, 0.0055], + device='cuda:1'), in_proj_covar=tensor([0.0016, 0.0016, 0.0016, 0.0012, 0.0011, 0.0015, 0.0010, 0.0011], + device='cuda:1'), out_proj_covar=tensor([1.1912e-05, 1.1516e-05, 1.2682e-05, 7.4984e-06, 6.9771e-06, 1.1490e-05, + 6.3568e-06, 6.8738e-06], device='cuda:1') +2023-03-20 17:34:38,084 INFO [train.py:901] (1/2) Epoch 1, batch 2050, loss[loss=0.3711, simple_loss=0.3805, pruned_loss=0.1809, over 7217.00 frames. ], tot_loss[loss=0.3702, simple_loss=0.3902, pruned_loss=0.19, over 1443495.15 frames. ], batch size: 50, lr: 4.81e-02, grad_scale: 16.0 +2023-03-20 17:34:38,109 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 17:34:56,925 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2090.0, num_to_drop=2, layers_to_drop={1, 2} +2023-03-20 17:35:01,814 INFO [train.py:901] (1/2) Epoch 1, batch 2100, loss[loss=0.386, simple_loss=0.3961, pruned_loss=0.188, over 7286.00 frames. ], tot_loss[loss=0.3696, simple_loss=0.3885, pruned_loss=0.1869, over 1444085.74 frames. ], batch size: 68, lr: 4.80e-02, grad_scale: 16.0 +2023-03-20 17:35:02,252 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.029e+02 3.619e+02 4.625e+02 6.076e+02 9.529e+02, threshold=9.250e+02, percent-clipped=0.0 +2023-03-20 17:35:08,988 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 17:35:12,369 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 17:35:18,921 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.36 vs. limit=5.0 +2023-03-20 17:35:19,195 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=2138.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:35:25,653 INFO [train.py:901] (1/2) Epoch 1, batch 2150, loss[loss=0.3348, simple_loss=0.3573, pruned_loss=0.1561, over 7347.00 frames. ], tot_loss[loss=0.366, simple_loss=0.3846, pruned_loss=0.1827, over 1445507.39 frames. ], batch size: 63, lr: 4.79e-02, grad_scale: 16.0 +2023-03-20 17:35:38,274 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6395, 3.6456, 3.8049, 3.7836, 3.6669, 2.7486, 3.8266, 3.1850], + device='cuda:1'), covar=tensor([0.0446, 0.0447, 0.0307, 0.0579, 0.0533, 0.3849, 0.0366, 0.3175], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0019, 0.0023, 0.0024, 0.0024, 0.0042, 0.0022, 0.0037], + device='cuda:1'), out_proj_covar=tensor([1.2173e-05, 1.0830e-05, 1.3679e-05, 1.5484e-05, 1.5219e-05, 3.2922e-05, + 1.2840e-05, 3.0896e-05], device='cuda:1') +2023-03-20 17:35:45,424 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2194.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:35:49,114 INFO [train.py:901] (1/2) Epoch 1, batch 2200, loss[loss=0.3977, simple_loss=0.4039, pruned_loss=0.1958, over 7309.00 frames. ], tot_loss[loss=0.365, simple_loss=0.3833, pruned_loss=0.1803, over 1447021.09 frames. ], batch size: 83, lr: 4.78e-02, grad_scale: 16.0 +2023-03-20 17:35:49,561 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.340e+02 3.853e+02 4.847e+02 6.366e+02 1.558e+03, threshold=9.694e+02, percent-clipped=3.0 +2023-03-20 17:35:52,377 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 17:35:55,389 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2213.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:35:59,521 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2222.0, num_to_drop=2, layers_to_drop={1, 2} +2023-03-20 17:36:01,342 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2226.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:36:13,524 INFO [train.py:901] (1/2) Epoch 1, batch 2250, loss[loss=0.4022, simple_loss=0.4077, pruned_loss=0.1984, over 7205.00 frames. ], tot_loss[loss=0.3629, simple_loss=0.3817, pruned_loss=0.1774, over 1446673.17 frames. ], batch size: 93, lr: 4.77e-02, grad_scale: 16.0 +2023-03-20 17:36:21,032 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2267.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 17:36:22,422 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=2270.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:36:25,105 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 17:36:25,548 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 17:36:29,947 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5524, 3.6488, 3.1524, 3.6117, 3.8376, 3.3750, 3.5844, 3.2857], + device='cuda:1'), covar=tensor([0.0092, 0.0211, 0.0188, 0.0196, 0.0157, 0.0274, 0.0192, 0.0267], + device='cuda:1'), in_proj_covar=tensor([0.0013, 0.0016, 0.0016, 0.0017, 0.0015, 0.0018, 0.0017, 0.0018], + device='cuda:1'), out_proj_covar=tensor([9.8155e-06, 1.4184e-05, 1.3015e-05, 1.5802e-05, 1.2547e-05, 1.6352e-05, + 1.5002e-05, 1.6193e-05], device='cuda:1') +2023-03-20 17:36:36,312 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 17:36:36,771 INFO [train.py:901] (1/2) Epoch 1, batch 2300, loss[loss=0.3362, simple_loss=0.3702, pruned_loss=0.1511, over 7348.00 frames. ], tot_loss[loss=0.3594, simple_loss=0.3785, pruned_loss=0.1743, over 1444998.88 frames. ], batch size: 63, lr: 4.77e-02, grad_scale: 16.0 +2023-03-20 17:36:37,861 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.965e+02 3.824e+02 4.896e+02 6.606e+02 1.192e+03, threshold=9.792e+02, percent-clipped=5.0 +2023-03-20 17:36:44,048 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=2315.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:36:56,607 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2340.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:37:01,660 INFO [train.py:901] (1/2) Epoch 1, batch 2350, loss[loss=0.3282, simple_loss=0.3563, pruned_loss=0.15, over 7244.00 frames. ], tot_loss[loss=0.3564, simple_loss=0.3764, pruned_loss=0.1715, over 1444533.21 frames. ], batch size: 55, lr: 4.76e-02, grad_scale: 16.0 +2023-03-20 17:37:10,586 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2596, 3.6725, 3.6126, 3.4725, 3.2599, 3.6626, 4.0322, 3.9448], + device='cuda:1'), covar=tensor([0.0484, 0.0286, 0.0382, 0.0393, 0.0589, 0.0416, 0.0323, 0.0302], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0049, 0.0050, 0.0060, 0.0052, 0.0047, 0.0042, 0.0048], + device='cuda:1'), out_proj_covar=tensor([4.6157e-05, 4.2400e-05, 4.8642e-05, 5.6346e-05, 4.6809e-05, 4.3704e-05, + 3.8013e-05, 4.6009e-05], device='cuda:1') +2023-03-20 17:37:18,863 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-20 17:37:19,485 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 17:37:25,982 INFO [train.py:901] (1/2) Epoch 1, batch 2400, loss[loss=0.3337, simple_loss=0.3691, pruned_loss=0.1491, over 7350.00 frames. ], tot_loss[loss=0.3549, simple_loss=0.3751, pruned_loss=0.1699, over 1445907.26 frames. ], batch size: 73, lr: 4.75e-02, grad_scale: 16.0 +2023-03-20 17:37:25,992 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 17:37:26,128 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2401.0, num_to_drop=2, layers_to_drop={0, 1} +2023-03-20 17:37:26,429 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.851e+02 3.554e+02 4.568e+02 5.565e+02 8.265e+02, threshold=9.137e+02, percent-clipped=0.0 +2023-03-20 17:37:36,036 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 17:37:38,773 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 17:37:39,071 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 +2023-03-20 17:37:49,955 INFO [train.py:901] (1/2) Epoch 1, batch 2450, loss[loss=0.3458, simple_loss=0.3606, pruned_loss=0.1655, over 7327.00 frames. ], tot_loss[loss=0.3517, simple_loss=0.3719, pruned_loss=0.1677, over 1444178.44 frames. ], batch size: 44, lr: 4.74e-02, grad_scale: 16.0 +2023-03-20 17:37:58,757 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.39 vs. limit=2.0 +2023-03-20 17:37:58,762 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.16 vs. limit=5.0 +2023-03-20 17:38:04,304 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 17:38:11,560 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2494.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 17:38:14,769 INFO [train.py:901] (1/2) Epoch 1, batch 2500, loss[loss=0.3295, simple_loss=0.358, pruned_loss=0.1505, over 7338.00 frames. ], tot_loss[loss=0.3473, simple_loss=0.3687, pruned_loss=0.1645, over 1446370.15 frames. ], batch size: 61, lr: 4.73e-02, grad_scale: 16.0 +2023-03-20 17:38:15,214 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.011e+02 3.688e+02 4.749e+02 6.089e+02 1.180e+03, threshold=9.499e+02, percent-clipped=5.0 +2023-03-20 17:38:20,391 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2513.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:38:26,491 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2526.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:38:27,829 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 17:38:34,027 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=2542.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:38:34,789 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.09 vs. limit=2.0 +2023-03-20 17:38:37,527 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.15 vs. limit=2.0 +2023-03-20 17:38:38,171 INFO [train.py:901] (1/2) Epoch 1, batch 2550, loss[loss=0.3075, simple_loss=0.3312, pruned_loss=0.1419, over 7116.00 frames. ], tot_loss[loss=0.3442, simple_loss=0.3663, pruned_loss=0.1622, over 1446481.94 frames. ], batch size: 41, lr: 4.72e-02, grad_scale: 16.0 +2023-03-20 17:38:43,517 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=2561.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:38:44,044 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5747, 3.6475, 3.4838, 3.2908, 3.5404, 3.1407, 3.4255, 3.3378], + device='cuda:1'), covar=tensor([0.0132, 0.0149, 0.0139, 0.0289, 0.0241, 0.0292, 0.0268, 0.0256], + device='cuda:1'), in_proj_covar=tensor([0.0012, 0.0014, 0.0014, 0.0016, 0.0014, 0.0015, 0.0016, 0.0017], + device='cuda:1'), out_proj_covar=tensor([9.4779e-06, 1.2963e-05, 1.1899e-05, 1.5715e-05, 1.2177e-05, 1.4256e-05, + 1.4904e-05, 1.7175e-05], device='cuda:1') +2023-03-20 17:38:49,574 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=2574.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:39:03,551 INFO [train.py:901] (1/2) Epoch 1, batch 2600, loss[loss=0.286, simple_loss=0.3175, pruned_loss=0.1273, over 7340.00 frames. ], tot_loss[loss=0.3419, simple_loss=0.3646, pruned_loss=0.1605, over 1448412.24 frames. ], batch size: 44, lr: 4.71e-02, grad_scale: 16.0 +2023-03-20 17:39:04,020 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.100e+02 3.577e+02 4.452e+02 5.636e+02 1.247e+03, threshold=8.904e+02, percent-clipped=2.0 +2023-03-20 17:39:07,202 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.08 vs. limit=2.0 +2023-03-20 17:39:26,905 INFO [train.py:901] (1/2) Epoch 1, batch 2650, loss[loss=0.3372, simple_loss=0.3669, pruned_loss=0.1537, over 7364.00 frames. ], tot_loss[loss=0.3378, simple_loss=0.3614, pruned_loss=0.1579, over 1445926.13 frames. ], batch size: 73, lr: 4.70e-02, grad_scale: 16.0 +2023-03-20 17:39:41,453 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 +2023-03-20 17:39:43,067 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2685.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:39:48,173 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2696.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:39:50,387 INFO [train.py:901] (1/2) Epoch 1, batch 2700, loss[loss=0.3616, simple_loss=0.3768, pruned_loss=0.1733, over 7292.00 frames. ], tot_loss[loss=0.3361, simple_loss=0.3599, pruned_loss=0.1568, over 1442189.30 frames. ], batch size: 86, lr: 4.69e-02, grad_scale: 16.0 +2023-03-20 17:39:50,815 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.158e+02 3.938e+02 4.931e+02 6.424e+02 1.116e+03, threshold=9.861e+02, percent-clipped=3.0 +2023-03-20 17:40:05,986 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2724, 3.8763, 3.7309, 3.3509, 3.1723, 3.5721, 4.0420, 3.9336], + device='cuda:1'), covar=tensor([0.0651, 0.0263, 0.0422, 0.0571, 0.0870, 0.0579, 0.0367, 0.0413], + device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0052, 0.0052, 0.0067, 0.0055, 0.0053, 0.0045, 0.0052], + device='cuda:1'), out_proj_covar=tensor([5.5830e-05, 4.7765e-05, 5.4138e-05, 6.7612e-05, 5.3831e-05, 5.4152e-05, + 4.4201e-05, 5.1925e-05], device='cuda:1') +2023-03-20 17:40:11,672 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2746.0, num_to_drop=2, layers_to_drop={2, 3} +2023-03-20 17:40:12,236 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.02 vs. limit=2.0 +2023-03-20 17:40:13,953 INFO [train.py:901] (1/2) Epoch 1, batch 2750, loss[loss=0.3296, simple_loss=0.3689, pruned_loss=0.1452, over 7290.00 frames. ], tot_loss[loss=0.3359, simple_loss=0.3602, pruned_loss=0.1562, over 1441543.26 frames. ], batch size: 68, lr: 4.68e-02, grad_scale: 16.0 +2023-03-20 17:40:15,034 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2848, 3.1466, 3.4633, 3.4693, 3.0724, 2.5485, 3.5419, 2.9036], + device='cuda:1'), covar=tensor([0.0479, 0.0477, 0.0453, 0.0365, 0.0475, 0.3659, 0.0294, 0.2738], + device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0030, 0.0034, 0.0033, 0.0035, 0.0072, 0.0031, 0.0060], + device='cuda:1'), out_proj_covar=tensor([1.6524e-05, 1.4867e-05, 1.8141e-05, 2.0462e-05, 1.9879e-05, 5.2235e-05, + 1.6604e-05, 4.6448e-05], device='cuda:1') +2023-03-20 17:40:16,390 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3436, 2.5615, 2.5601, 2.2118, 2.0685, 2.2204, 2.2993, 2.6409], + device='cuda:1'), covar=tensor([0.0300, 0.0231, 0.0178, 0.0393, 0.0790, 0.0457, 0.0386, 0.0129], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0033, 0.0034, 0.0035, 0.0034, 0.0035, 0.0037, 0.0027], + device='cuda:1'), out_proj_covar=tensor([2.8947e-05, 2.4695e-05, 2.5906e-05, 2.8650e-05, 2.6214e-05, 2.8925e-05, + 3.0987e-05, 2.0050e-05], device='cuda:1') +2023-03-20 17:40:36,828 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 +2023-03-20 17:40:37,982 INFO [train.py:901] (1/2) Epoch 1, batch 2800, loss[loss=0.3334, simple_loss=0.369, pruned_loss=0.1489, over 7351.00 frames. ], tot_loss[loss=0.3371, simple_loss=0.3608, pruned_loss=0.157, over 1440919.32 frames. ], batch size: 54, lr: 4.67e-02, grad_scale: 16.0 +2023-03-20 17:40:38,406 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.290e+02 4.051e+02 5.024e+02 6.754e+02 1.289e+03, threshold=1.005e+03, percent-clipped=6.0 +2023-03-20 17:41:02,445 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 17:41:02,775 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 17:41:02,798 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 17:41:02,815 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 17:41:02,981 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 17:41:02,996 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 17:41:03,007 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 17:41:03,103 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 17:41:03,177 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 17:41:03,191 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 17:41:03,208 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 17:41:03,365 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 17:41:03,371 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 17:41:03,379 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 17:41:03,808 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 17:41:03,812 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 17:41:03,868 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 17:41:10,781 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2824.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:41:11,121 INFO [train.py:901] (1/2) Epoch 2, batch 0, loss[loss=0.3263, simple_loss=0.3504, pruned_loss=0.1511, over 7246.00 frames. ], tot_loss[loss=0.3263, simple_loss=0.3504, pruned_loss=0.1511, over 7246.00 frames. ], batch size: 55, lr: 4.63e-02, grad_scale: 16.0 +2023-03-20 17:41:11,121 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 17:41:14,724 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8134, 3.4111, 3.3244, 3.0047, 2.7826, 2.4941, 2.8426, 3.4780], + device='cuda:1'), covar=tensor([0.0762, 0.0166, 0.0267, 0.0371, 0.0655, 0.1141, 0.0601, 0.0134], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0020, 0.0021, 0.0023, 0.0026, 0.0024, 0.0025, 0.0020], + device='cuda:1'), out_proj_covar=tensor([1.4233e-05, 1.3068e-05, 1.4441e-05, 1.6323e-05, 2.0179e-05, 1.8408e-05, + 1.7306e-05, 1.2943e-05], device='cuda:1') +2023-03-20 17:41:21,852 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3110, 3.5353, 3.2753, 3.2928, 3.4862, 3.2402, 3.3984, 3.0774], + device='cuda:1'), covar=tensor([0.0134, 0.0155, 0.0151, 0.0186, 0.0186, 0.0201, 0.0212, 0.0291], + device='cuda:1'), in_proj_covar=tensor([0.0011, 0.0013, 0.0014, 0.0014, 0.0012, 0.0014, 0.0015, 0.0015], + device='cuda:1'), out_proj_covar=tensor([9.1722e-06, 1.2904e-05, 1.1639e-05, 1.4479e-05, 1.1425e-05, 1.3612e-05, + 1.4300e-05, 1.5178e-05], device='cuda:1') +2023-03-20 17:41:36,988 INFO [train.py:935] (1/2) Epoch 2, validation: loss=0.3661, simple_loss=0.5483, pruned_loss=0.09198, over 1622729.00 frames. +2023-03-20 17:41:36,989 INFO [train.py:936] (1/2) Maximum memory allocated so far is 11314MB +2023-03-20 17:41:40,034 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.33 vs. limit=2.0 +2023-03-20 17:41:43,581 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 17:41:51,786 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-20 17:41:52,963 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 17:41:59,964 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 17:42:00,428 INFO [train.py:901] (1/2) Epoch 2, batch 50, loss[loss=0.288, simple_loss=0.3315, pruned_loss=0.1222, over 7345.00 frames. ], tot_loss[loss=0.3316, simple_loss=0.3567, pruned_loss=0.1532, over 325888.91 frames. ], batch size: 44, lr: 4.62e-02, grad_scale: 16.0 +2023-03-20 17:42:02,348 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 17:42:04,773 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 17:42:05,396 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2885.0, num_to_drop=2, layers_to_drop={0, 2} +2023-03-20 17:42:13,253 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.499e+02 3.706e+02 4.761e+02 5.953e+02 9.825e+02, threshold=9.522e+02, percent-clipped=0.0 +2023-03-20 17:42:22,607 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 17:42:23,086 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 17:42:25,469 INFO [train.py:901] (1/2) Epoch 2, batch 100, loss[loss=0.3204, simple_loss=0.3482, pruned_loss=0.1463, over 7320.00 frames. ], tot_loss[loss=0.3309, simple_loss=0.3573, pruned_loss=0.1522, over 575651.61 frames. ], batch size: 80, lr: 4.61e-02, grad_scale: 16.0 +2023-03-20 17:42:31,299 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2937.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:42:49,393 INFO [train.py:901] (1/2) Epoch 2, batch 150, loss[loss=0.3436, simple_loss=0.363, pruned_loss=0.1622, over 7284.00 frames. ], tot_loss[loss=0.326, simple_loss=0.3536, pruned_loss=0.1492, over 769199.82 frames. ], batch size: 77, lr: 4.60e-02, grad_scale: 16.0 +2023-03-20 17:42:59,680 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2996.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 17:43:01,299 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2998.0, num_to_drop=2, layers_to_drop={2, 3} +2023-03-20 17:43:03,380 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.879e+02 3.800e+02 4.704e+02 5.909e+02 1.183e+03, threshold=9.409e+02, percent-clipped=3.0 +2023-03-20 17:43:11,810 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3019.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:43:14,527 INFO [train.py:901] (1/2) Epoch 2, batch 200, loss[loss=0.3279, simple_loss=0.3616, pruned_loss=0.1472, over 7281.00 frames. ], tot_loss[loss=0.3254, simple_loss=0.3533, pruned_loss=0.1487, over 919599.93 frames. ], batch size: 77, lr: 4.59e-02, grad_scale: 16.0 +2023-03-20 17:43:19,424 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 17:43:22,465 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3041.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:43:23,896 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 17:43:23,940 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=3044.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:43:29,207 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 17:43:33,039 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-20 17:43:38,975 INFO [train.py:901] (1/2) Epoch 2, batch 250, loss[loss=0.312, simple_loss=0.3478, pruned_loss=0.1381, over 7283.00 frames. ], tot_loss[loss=0.3241, simple_loss=0.3528, pruned_loss=0.1477, over 1035742.86 frames. ], batch size: 66, lr: 4.58e-02, grad_scale: 16.0 +2023-03-20 17:43:41,971 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 17:43:42,096 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3080.0, num_to_drop=2, layers_to_drop={1, 2} +2023-03-20 17:43:42,816 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-20 17:43:52,950 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.532e+02 4.005e+02 4.790e+02 6.030e+02 1.284e+03, threshold=9.580e+02, percent-clipped=2.0 +2023-03-20 17:44:01,247 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 17:44:04,151 INFO [train.py:901] (1/2) Epoch 2, batch 300, loss[loss=0.3275, simple_loss=0.3571, pruned_loss=0.1489, over 7171.00 frames. ], tot_loss[loss=0.3224, simple_loss=0.3514, pruned_loss=0.1468, over 1124329.06 frames. ], batch size: 98, lr: 4.57e-02, grad_scale: 16.0 +2023-03-20 17:44:10,548 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 17:44:29,659 INFO [train.py:901] (1/2) Epoch 2, batch 350, loss[loss=0.3516, simple_loss=0.3688, pruned_loss=0.1672, over 7311.00 frames. ], tot_loss[loss=0.3203, simple_loss=0.3495, pruned_loss=0.1455, over 1193963.73 frames. ], batch size: 83, lr: 4.56e-02, grad_scale: 16.0 +2023-03-20 17:44:29,801 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0382, 3.2895, 3.0212, 2.9089, 2.8855, 2.8271, 2.5488, 2.9416], + device='cuda:1'), covar=tensor([0.0220, 0.0143, 0.0196, 0.0268, 0.0217, 0.0245, 0.0394, 0.0170], + device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0030, 0.0032, 0.0033, 0.0031, 0.0032, 0.0034, 0.0025], + device='cuda:1'), out_proj_covar=tensor([2.7829e-05, 2.2429e-05, 2.4862e-05, 2.6528e-05, 2.3095e-05, 2.6284e-05, + 2.8285e-05, 1.9138e-05], device='cuda:1') +2023-03-20 17:44:32,102 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3180.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:44:42,793 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.966e+02 3.735e+02 4.987e+02 6.061e+02 9.476e+02, threshold=9.975e+02, percent-clipped=1.0 +2023-03-20 17:44:45,143 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 17:44:53,759 INFO [train.py:901] (1/2) Epoch 2, batch 400, loss[loss=0.2502, simple_loss=0.2832, pruned_loss=0.1087, over 7003.00 frames. ], tot_loss[loss=0.3196, simple_loss=0.3488, pruned_loss=0.1452, over 1250467.01 frames. ], batch size: 35, lr: 4.55e-02, grad_scale: 16.0 +2023-03-20 17:44:58,600 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3412, 2.1413, 3.5066, 3.2088, 3.5054, 3.3454, 2.4964, 3.2744], + device='cuda:1'), covar=tensor([0.0192, 0.0793, 0.0077, 0.0209, 0.0090, 0.0203, 0.1144, 0.0151], + device='cuda:1'), in_proj_covar=tensor([0.0013, 0.0016, 0.0014, 0.0014, 0.0014, 0.0016, 0.0025, 0.0014], + device='cuda:1'), out_proj_covar=tensor([1.3149e-05, 1.7242e-05, 1.4627e-05, 1.3946e-05, 1.1978e-05, 1.4161e-05, + 2.8603e-05, 1.4313e-05], device='cuda:1') +2023-03-20 17:45:08,681 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.17 vs. limit=2.0 +2023-03-20 17:45:18,910 INFO [train.py:901] (1/2) Epoch 2, batch 450, loss[loss=0.2739, simple_loss=0.3183, pruned_loss=0.1148, over 7277.00 frames. ], tot_loss[loss=0.3171, simple_loss=0.3465, pruned_loss=0.1438, over 1293572.15 frames. ], batch size: 47, lr: 4.54e-02, grad_scale: 16.0 +2023-03-20 17:45:25,256 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3288.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:45:26,084 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 17:45:26,546 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 17:45:27,615 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3293.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:45:31,876 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.013e+02 3.842e+02 4.912e+02 6.234e+02 9.651e+02, threshold=9.824e+02, percent-clipped=0.0 +2023-03-20 17:45:42,135 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.9206, 4.9721, 4.8123, 5.1615, 4.8470, 4.6214, 5.2826, 5.0911], + device='cuda:1'), covar=tensor([0.0313, 0.0193, 0.0289, 0.0306, 0.0267, 0.0320, 0.0207, 0.0208], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0040, 0.0044, 0.0040, 0.0037, 0.0047, 0.0039, 0.0044], + device='cuda:1'), out_proj_covar=tensor([4.7345e-05, 3.9228e-05, 4.7576e-05, 4.3659e-05, 3.7056e-05, 5.0979e-05, + 4.0241e-05, 4.8852e-05], device='cuda:1') +2023-03-20 17:45:43,012 INFO [train.py:901] (1/2) Epoch 2, batch 500, loss[loss=0.2865, simple_loss=0.3245, pruned_loss=0.1243, over 7138.00 frames. ], tot_loss[loss=0.3155, simple_loss=0.3452, pruned_loss=0.1428, over 1327235.56 frames. ], batch size: 41, lr: 4.53e-02, grad_scale: 16.0 +2023-03-20 17:45:45,575 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6618, 3.0250, 3.2935, 2.7025, 2.9361, 2.1411, 3.5668, 3.0172], + device='cuda:1'), covar=tensor([0.0473, 0.0184, 0.0218, 0.0772, 0.0447, 0.1065, 0.0097, 0.0445], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0024, 0.0023, 0.0029, 0.0026, 0.0025, 0.0020, 0.0029], + device='cuda:1'), out_proj_covar=tensor([2.3172e-05, 1.8655e-05, 1.8192e-05, 2.6389e-05, 2.2678e-05, 2.1674e-05, + 1.6022e-05, 2.6553e-05], device='cuda:1') +2023-03-20 17:45:49,710 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.01 vs. limit=2.0 +2023-03-20 17:45:50,931 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3341.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:45:55,461 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3349.0, num_to_drop=2, layers_to_drop={0, 1} +2023-03-20 17:45:59,357 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 17:46:00,372 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 17:46:00,849 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 17:46:02,784 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 17:46:07,513 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 17:46:08,469 INFO [train.py:901] (1/2) Epoch 2, batch 550, loss[loss=0.3106, simple_loss=0.3455, pruned_loss=0.1379, over 7270.00 frames. ], tot_loss[loss=0.3143, simple_loss=0.3441, pruned_loss=0.1422, over 1351463.13 frames. ], batch size: 89, lr: 4.52e-02, grad_scale: 16.0 +2023-03-20 17:46:08,544 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3375.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:46:09,591 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8882, 3.2685, 3.1055, 3.1563, 2.4347, 2.2838, 2.7407, 2.9533], + device='cuda:1'), covar=tensor([0.0675, 0.0174, 0.0175, 0.0153, 0.0457, 0.0635, 0.0463, 0.0212], + device='cuda:1'), in_proj_covar=tensor([0.0022, 0.0022, 0.0022, 0.0022, 0.0024, 0.0022, 0.0024, 0.0021], + device='cuda:1'), out_proj_covar=tensor([1.7520e-05, 1.6955e-05, 1.6939e-05, 1.8140e-05, 2.0982e-05, 1.9101e-05, + 1.9350e-05, 1.6051e-05], device='cuda:1') +2023-03-20 17:46:15,477 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=3389.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:46:17,867 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 17:46:22,174 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.129e+02 3.792e+02 4.639e+02 5.385e+02 1.138e+03, threshold=9.277e+02, percent-clipped=1.0 +2023-03-20 17:46:25,099 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 17:46:28,078 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 17:46:33,410 INFO [train.py:901] (1/2) Epoch 2, batch 600, loss[loss=0.3725, simple_loss=0.3875, pruned_loss=0.1787, over 7344.00 frames. ], tot_loss[loss=0.3154, simple_loss=0.345, pruned_loss=0.1429, over 1371426.24 frames. ], batch size: 54, lr: 4.51e-02, grad_scale: 8.0 +2023-03-20 17:46:34,802 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-03-20 17:46:34,871 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 17:46:51,286 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 17:46:59,228 INFO [train.py:901] (1/2) Epoch 2, batch 650, loss[loss=0.3564, simple_loss=0.3758, pruned_loss=0.1685, over 7361.00 frames. ], tot_loss[loss=0.3141, simple_loss=0.3439, pruned_loss=0.1421, over 1386884.32 frames. ], batch size: 63, lr: 4.50e-02, grad_scale: 8.0 +2023-03-20 17:47:01,105 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 17:47:01,663 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3480.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:47:04,583 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5991, 4.6752, 4.5993, 4.9399, 5.0639, 4.9333, 4.1579, 4.2576], + device='cuda:1'), covar=tensor([0.0843, 0.0732, 0.0820, 0.1186, 0.0508, 0.0742, 0.0633, 0.0761], + device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0094, 0.0108, 0.0083, 0.0079, 0.0086, 0.0064, 0.0078], + device='cuda:1'), out_proj_covar=tensor([6.4986e-05, 9.8308e-05, 1.2042e-04, 9.1338e-05, 8.0501e-05, 8.9441e-05, + 6.4906e-05, 7.6549e-05], device='cuda:1') +2023-03-20 17:47:12,719 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.910e+02 3.942e+02 5.016e+02 7.183e+02 2.296e+03, threshold=1.003e+03, percent-clipped=6.0 +2023-03-20 17:47:17,268 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 17:47:24,894 INFO [train.py:901] (1/2) Epoch 2, batch 700, loss[loss=0.3322, simple_loss=0.3611, pruned_loss=0.1517, over 7290.00 frames. ], tot_loss[loss=0.3135, simple_loss=0.3438, pruned_loss=0.1417, over 1398217.37 frames. ], batch size: 77, lr: 4.49e-02, grad_scale: 8.0 +2023-03-20 17:47:26,469 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=3528.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:47:26,888 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 17:47:28,069 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3920, 2.5064, 2.7753, 2.4747, 2.6688, 2.9752, 2.4411, 2.2890], + device='cuda:1'), covar=tensor([0.0210, 0.0302, 0.1246, 0.0612, 0.0201, 0.0267, 0.1531, 0.1469], + device='cuda:1'), in_proj_covar=tensor([0.0023, 0.0030, 0.0044, 0.0029, 0.0027, 0.0026, 0.0054, 0.0045], + device='cuda:1'), out_proj_covar=tensor([1.2854e-05, 1.7414e-05, 3.4089e-05, 1.7845e-05, 1.5365e-05, 1.4783e-05, + 4.4083e-05, 3.1386e-05], device='cuda:1') +2023-03-20 17:47:28,715 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-03-20 17:47:35,274 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3546.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:47:49,464 INFO [train.py:901] (1/2) Epoch 2, batch 750, loss[loss=0.3018, simple_loss=0.3302, pruned_loss=0.1367, over 7239.00 frames. ], tot_loss[loss=0.3122, simple_loss=0.3433, pruned_loss=0.1405, over 1410283.21 frames. ], batch size: 55, lr: 4.48e-02, grad_scale: 8.0 +2023-03-20 17:47:49,984 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 17:47:50,929 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 17:47:51,982 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5967, 3.7049, 3.6927, 3.4683, 3.5484, 3.8566, 4.1488, 3.6490], + device='cuda:1'), covar=tensor([0.0227, 0.0190, 0.0152, 0.0245, 0.0207, 0.0157, 0.0141, 0.0259], + device='cuda:1'), in_proj_covar=tensor([0.0026, 0.0024, 0.0027, 0.0024, 0.0030, 0.0026, 0.0023, 0.0024], + device='cuda:1'), out_proj_covar=tensor([3.0346e-05, 2.4169e-05, 3.3266e-05, 3.0725e-05, 3.5940e-05, 2.6473e-05, + 2.8313e-05, 2.7371e-05], device='cuda:1') +2023-03-20 17:47:57,487 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7427, 3.5561, 3.5970, 3.6120, 3.3623, 2.8418, 3.4682, 2.4966], + device='cuda:1'), covar=tensor([0.0425, 0.0537, 0.0366, 0.0406, 0.0313, 0.1013, 0.0254, 0.1094], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0023, 0.0022, 0.0022, 0.0023, 0.0023, 0.0026, 0.0022], + device='cuda:1'), out_proj_covar=tensor([1.8292e-05, 1.7909e-05, 1.7444e-05, 1.8663e-05, 2.0148e-05, 2.0183e-05, + 2.1856e-05, 1.7261e-05], device='cuda:1') +2023-03-20 17:47:58,449 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3593.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:48:00,150 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.12 vs. limit=5.0 +2023-03-20 17:48:03,622 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.081e+02 4.174e+02 5.233e+02 6.266e+02 1.011e+03, threshold=1.047e+03, percent-clipped=1.0 +2023-03-20 17:48:05,255 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 17:48:06,421 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3607.0, num_to_drop=2, layers_to_drop={2, 3} +2023-03-20 17:48:08,102 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-03-20 17:48:10,801 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 17:48:15,800 INFO [train.py:901] (1/2) Epoch 2, batch 800, loss[loss=0.3203, simple_loss=0.3537, pruned_loss=0.1435, over 7343.00 frames. ], tot_loss[loss=0.311, simple_loss=0.3423, pruned_loss=0.1399, over 1415759.21 frames. ], batch size: 73, lr: 4.47e-02, grad_scale: 8.0 +2023-03-20 17:48:15,830 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 17:48:17,374 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 17:48:23,799 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=3641.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:48:25,360 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3644.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 17:48:28,204 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 17:48:40,929 INFO [train.py:901] (1/2) Epoch 2, batch 850, loss[loss=0.3262, simple_loss=0.365, pruned_loss=0.1437, over 7329.00 frames. ], tot_loss[loss=0.3087, simple_loss=0.3406, pruned_loss=0.1384, over 1422948.65 frames. ], batch size: 75, lr: 4.46e-02, grad_scale: 8.0 +2023-03-20 17:48:41,067 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3675.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:48:41,591 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3676.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:48:43,573 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1059, 3.9513, 3.9534, 3.4968, 3.9786, 4.2065, 4.3361, 3.9242], + device='cuda:1'), covar=tensor([0.0167, 0.0114, 0.0167, 0.0346, 0.0123, 0.0105, 0.0123, 0.0176], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0024, 0.0029, 0.0026, 0.0031, 0.0027, 0.0023, 0.0025], + device='cuda:1'), out_proj_covar=tensor([3.2076e-05, 2.5328e-05, 3.7158e-05, 3.3111e-05, 3.7378e-05, 2.8543e-05, + 2.8300e-05, 2.8428e-05], device='cuda:1') +2023-03-20 17:48:46,996 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 17:48:47,005 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 17:48:47,882 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.09 vs. limit=2.0 +2023-03-20 17:48:48,398 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.18 vs. limit=2.0 +2023-03-20 17:48:53,095 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 17:48:56,249 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 3.969e+02 4.736e+02 6.049e+02 1.256e+03, threshold=9.472e+02, percent-clipped=4.0 +2023-03-20 17:48:57,257 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 17:49:06,506 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=3723.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:49:07,473 INFO [train.py:901] (1/2) Epoch 2, batch 900, loss[loss=0.2847, simple_loss=0.3222, pruned_loss=0.1236, over 7322.00 frames. ], tot_loss[loss=0.3071, simple_loss=0.3397, pruned_loss=0.1373, over 1426974.80 frames. ], batch size: 49, lr: 4.45e-02, grad_scale: 8.0 +2023-03-20 17:49:13,772 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3737.0, num_to_drop=2, layers_to_drop={2, 3} +2023-03-20 17:49:32,674 INFO [train.py:901] (1/2) Epoch 2, batch 950, loss[loss=0.2897, simple_loss=0.3354, pruned_loss=0.122, over 7369.00 frames. ], tot_loss[loss=0.3069, simple_loss=0.3397, pruned_loss=0.1371, over 1430162.29 frames. ], batch size: 63, lr: 4.44e-02, grad_scale: 8.0 +2023-03-20 17:49:34,178 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 17:49:48,230 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.101e+02 3.953e+02 4.768e+02 6.187e+02 9.797e+02, threshold=9.537e+02, percent-clipped=2.0 +2023-03-20 17:49:59,330 INFO [train.py:901] (1/2) Epoch 2, batch 1000, loss[loss=0.2943, simple_loss=0.338, pruned_loss=0.1253, over 7287.00 frames. ], tot_loss[loss=0.3048, simple_loss=0.3382, pruned_loss=0.1357, over 1433406.90 frames. ], batch size: 66, lr: 4.43e-02, grad_scale: 8.0 +2023-03-20 17:49:59,344 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 17:50:05,495 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5259, 2.5844, 2.6191, 2.8030, 2.7961, 2.3057, 2.8847, 2.8456], + device='cuda:1'), covar=tensor([0.0958, 0.0681, 0.0905, 0.0737, 0.0983, 0.2863, 0.0726, 0.2320], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0034, 0.0039, 0.0043, 0.0042, 0.0086, 0.0039, 0.0072], + device='cuda:1'), out_proj_covar=tensor([1.8354e-05, 1.7176e-05, 2.1034e-05, 2.5399e-05, 2.1606e-05, 5.4264e-05, + 1.9376e-05, 4.9332e-05], device='cuda:1') +2023-03-20 17:50:10,282 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 +2023-03-20 17:50:20,146 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.01 vs. limit=2.0 +2023-03-20 17:50:20,335 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 17:50:25,599 INFO [train.py:901] (1/2) Epoch 2, batch 1050, loss[loss=0.3054, simple_loss=0.3311, pruned_loss=0.1399, over 7179.00 frames. ], tot_loss[loss=0.3055, simple_loss=0.3386, pruned_loss=0.1362, over 1433148.69 frames. ], batch size: 39, lr: 4.41e-02, grad_scale: 8.0 +2023-03-20 17:50:40,149 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3902.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:50:40,538 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.085e+02 3.872e+02 4.717e+02 6.455e+02 1.344e+03, threshold=9.435e+02, percent-clipped=6.0 +2023-03-20 17:50:43,037 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 17:50:47,653 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 17:50:51,701 INFO [train.py:901] (1/2) Epoch 2, batch 1100, loss[loss=0.2878, simple_loss=0.334, pruned_loss=0.1208, over 7318.00 frames. ], tot_loss[loss=0.3043, simple_loss=0.338, pruned_loss=0.1353, over 1434887.61 frames. ], batch size: 86, lr: 4.40e-02, grad_scale: 8.0 +2023-03-20 17:50:52,382 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3926.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:51:01,600 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3944.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:51:07,459 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.03 vs. limit=5.0 +2023-03-20 17:51:17,278 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 17:51:17,790 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 17:51:18,781 INFO [train.py:901] (1/2) Epoch 2, batch 1150, loss[loss=0.2555, simple_loss=0.2999, pruned_loss=0.1055, over 7248.00 frames. ], tot_loss[loss=0.3022, simple_loss=0.3363, pruned_loss=0.1341, over 1436807.70 frames. ], batch size: 47, lr: 4.39e-02, grad_scale: 8.0 +2023-03-20 17:51:25,052 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3987.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:51:27,232 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 +2023-03-20 17:51:27,461 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=3992.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 17:51:30,389 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 17:51:30,926 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 17:51:36,689 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.009e+02 4.029e+02 4.723e+02 6.247e+02 1.092e+03, threshold=9.446e+02, percent-clipped=3.0 +2023-03-20 17:51:39,466 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3467, 2.7004, 2.2997, 2.4279, 2.0460, 2.1047, 2.6835, 2.3071], + device='cuda:1'), covar=tensor([0.0509, 0.0239, 0.0334, 0.0191, 0.0476, 0.0612, 0.0343, 0.0365], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0027, 0.0028, 0.0027, 0.0031, 0.0028, 0.0032, 0.0027], + device='cuda:1'), out_proj_covar=tensor([2.6535e-05, 2.4048e-05, 2.4483e-05, 2.3928e-05, 2.8578e-05, 2.6870e-05, + 2.9321e-05, 2.3540e-05], device='cuda:1') +2023-03-20 17:51:48,208 INFO [train.py:901] (1/2) Epoch 2, batch 1200, loss[loss=0.211, simple_loss=0.2471, pruned_loss=0.08751, over 6049.00 frames. ], tot_loss[loss=0.301, simple_loss=0.3355, pruned_loss=0.1333, over 1438627.64 frames. ], batch size: 26, lr: 4.38e-02, grad_scale: 8.0 +2023-03-20 17:51:51,861 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4032.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:51:54,698 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.63 vs. limit=2.0 +2023-03-20 17:52:09,000 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 17:52:15,179 INFO [train.py:901] (1/2) Epoch 2, batch 1250, loss[loss=0.2961, simple_loss=0.3243, pruned_loss=0.1339, over 7286.00 frames. ], tot_loss[loss=0.3021, simple_loss=0.3364, pruned_loss=0.134, over 1438126.70 frames. ], batch size: 47, lr: 4.37e-02, grad_scale: 8.0 +2023-03-20 17:52:16,062 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.14 vs. limit=2.0 +2023-03-20 17:52:29,536 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.597e+02 4.003e+02 4.997e+02 6.171e+02 1.262e+03, threshold=9.993e+02, percent-clipped=4.0 +2023-03-20 17:52:32,154 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 17:52:34,500 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.27 vs. limit=2.0 +2023-03-20 17:52:36,249 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 17:52:37,356 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 17:52:39,532 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5027, 2.7814, 2.7521, 2.4559, 2.7554, 2.4193, 2.8682, 2.7988], + device='cuda:1'), covar=tensor([0.0362, 0.0066, 0.0197, 0.0446, 0.0246, 0.0454, 0.0104, 0.0169], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0024, 0.0029, 0.0031, 0.0032, 0.0029, 0.0024, 0.0034], + device='cuda:1'), out_proj_covar=tensor([2.9659e-05, 2.1087e-05, 2.7383e-05, 3.2836e-05, 3.2528e-05, 3.0045e-05, + 2.1910e-05, 3.6380e-05], device='cuda:1') +2023-03-20 17:52:40,929 INFO [train.py:901] (1/2) Epoch 2, batch 1300, loss[loss=0.3081, simple_loss=0.3445, pruned_loss=0.1358, over 7227.00 frames. ], tot_loss[loss=0.3008, simple_loss=0.3357, pruned_loss=0.1329, over 1437972.61 frames. ], batch size: 93, lr: 4.36e-02, grad_scale: 8.0 +2023-03-20 17:52:50,446 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4141.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:53:02,196 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 17:53:04,740 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 17:53:06,892 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0763, 3.9077, 3.8988, 3.5366, 3.8713, 4.0985, 4.3488, 4.1524], + device='cuda:1'), covar=tensor([0.0207, 0.0146, 0.0221, 0.0261, 0.0195, 0.0129, 0.0135, 0.0121], + device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0028, 0.0032, 0.0027, 0.0036, 0.0032, 0.0027, 0.0027], + device='cuda:1'), out_proj_covar=tensor([3.8515e-05, 3.1857e-05, 4.5640e-05, 3.8952e-05, 4.7268e-05, 3.7735e-05, + 3.5521e-05, 3.2610e-05], device='cuda:1') +2023-03-20 17:53:07,826 INFO [train.py:901] (1/2) Epoch 2, batch 1350, loss[loss=0.2708, simple_loss=0.3157, pruned_loss=0.113, over 7310.00 frames. ], tot_loss[loss=0.2996, simple_loss=0.3348, pruned_loss=0.1322, over 1439160.12 frames. ], batch size: 86, lr: 4.35e-02, grad_scale: 8.0 +2023-03-20 17:53:08,402 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 17:53:18,298 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1782, 3.4526, 3.4434, 3.2585, 3.2577, 3.1343, 3.5911, 3.6555], + device='cuda:1'), covar=tensor([0.0415, 0.0296, 0.0274, 0.0322, 0.0365, 0.0599, 0.0305, 0.0267], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0063, 0.0059, 0.0076, 0.0062, 0.0053, 0.0052, 0.0053], + device='cuda:1'), out_proj_covar=tensor([7.4534e-05, 7.7471e-05, 7.1141e-05, 1.0126e-04, 7.8529e-05, 6.5159e-05, + 6.1050e-05, 6.6357e-05], device='cuda:1') +2023-03-20 17:53:18,690 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 17:53:22,275 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4202.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 17:53:22,303 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4202.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:53:22,643 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.237e+02 3.810e+02 4.674e+02 6.452e+02 1.079e+03, threshold=9.348e+02, percent-clipped=2.0 +2023-03-20 17:53:26,533 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-20 17:53:34,472 INFO [train.py:901] (1/2) Epoch 2, batch 1400, loss[loss=0.2981, simple_loss=0.3464, pruned_loss=0.125, over 7306.00 frames. ], tot_loss[loss=0.299, simple_loss=0.3346, pruned_loss=0.1317, over 1436555.46 frames. ], batch size: 68, lr: 4.34e-02, grad_scale: 8.0 +2023-03-20 17:53:40,381 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6559, 1.5913, 2.5075, 1.9383, 2.1175, 1.9155, 2.1122, 2.2980], + device='cuda:1'), covar=tensor([0.0820, 0.0771, 0.0336, 0.0567, 0.0635, 0.0865, 0.0392, 0.0221], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0037, 0.0040, 0.0046, 0.0039, 0.0039, 0.0038, 0.0036], + device='cuda:1'), out_proj_covar=tensor([4.1275e-05, 3.7473e-05, 4.4806e-05, 4.5530e-05, 3.9345e-05, 3.8493e-05, + 3.4780e-05, 3.0304e-05], device='cuda:1') +2023-03-20 17:53:48,082 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=4250.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:53:51,572 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 17:53:54,479 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=6.44 vs. limit=5.0 +2023-03-20 17:53:55,348 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3496, 3.4228, 3.4054, 3.3096, 3.2596, 2.1654, 3.2629, 3.2596], + device='cuda:1'), covar=tensor([0.0336, 0.0174, 0.0251, 0.0416, 0.0392, 0.2055, 0.0277, 0.1437], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0032, 0.0038, 0.0041, 0.0041, 0.0087, 0.0039, 0.0074], + device='cuda:1'), out_proj_covar=tensor([1.8577e-05, 1.6520e-05, 2.0158e-05, 2.4091e-05, 2.2074e-05, 5.5282e-05, + 2.0875e-05, 5.0191e-05], device='cuda:1') +2023-03-20 17:54:00,756 INFO [train.py:901] (1/2) Epoch 2, batch 1450, loss[loss=0.3169, simple_loss=0.345, pruned_loss=0.1444, over 7297.00 frames. ], tot_loss[loss=0.3004, simple_loss=0.3357, pruned_loss=0.1325, over 1436057.33 frames. ], batch size: 77, lr: 4.33e-02, grad_scale: 8.0 +2023-03-20 17:54:04,451 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4282.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:54:07,602 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4288.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:54:14,270 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 17:54:15,760 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.055e+02 3.966e+02 5.157e+02 6.398e+02 1.604e+03, threshold=1.031e+03, percent-clipped=6.0 +2023-03-20 17:54:20,348 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.13 vs. limit=2.0 +2023-03-20 17:54:27,659 INFO [train.py:901] (1/2) Epoch 2, batch 1500, loss[loss=0.2776, simple_loss=0.3204, pruned_loss=0.1174, over 7282.00 frames. ], tot_loss[loss=0.2983, simple_loss=0.3341, pruned_loss=0.1312, over 1438302.32 frames. ], batch size: 86, lr: 4.32e-02, grad_scale: 8.0 +2023-03-20 17:54:31,339 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 17:54:31,453 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4332.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 17:54:40,221 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4349.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:54:48,215 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.17 vs. limit=5.0 +2023-03-20 17:54:49,543 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5131, 2.3653, 2.1063, 2.4212, 1.7883, 2.1839, 2.3507, 1.5457], + device='cuda:1'), covar=tensor([0.0502, 0.0565, 0.0659, 0.0359, 0.0686, 0.0632, 0.0933, 0.0957], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0031, 0.0030, 0.0031, 0.0034, 0.0029, 0.0035, 0.0029], + device='cuda:1'), out_proj_covar=tensor([2.9865e-05, 2.9742e-05, 2.7932e-05, 2.7583e-05, 3.3667e-05, 2.9717e-05, + 3.4803e-05, 2.7617e-05], device='cuda:1') +2023-03-20 17:54:53,521 INFO [train.py:901] (1/2) Epoch 2, batch 1550, loss[loss=0.3083, simple_loss=0.3473, pruned_loss=0.1346, over 7341.00 frames. ], tot_loss[loss=0.2965, simple_loss=0.3327, pruned_loss=0.1301, over 1437900.20 frames. ], batch size: 54, lr: 4.31e-02, grad_scale: 8.0 +2023-03-20 17:54:56,632 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 17:54:56,693 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=4380.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 17:55:09,256 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.966e+02 3.883e+02 4.668e+02 5.736e+02 1.136e+03, threshold=9.337e+02, percent-clipped=1.0 +2023-03-20 17:55:17,031 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6486, 2.4469, 1.7036, 2.2873, 2.6161, 2.7387, 2.7373, 3.0615], + device='cuda:1'), covar=tensor([0.0286, 0.0082, 0.0837, 0.0472, 0.0401, 0.0205, 0.0162, 0.0106], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0021, 0.0026, 0.0026, 0.0028, 0.0022, 0.0022, 0.0027], + device='cuda:1'), out_proj_covar=tensor([2.4875e-05, 1.8860e-05, 2.6788e-05, 2.8511e-05, 3.0546e-05, 2.3483e-05, + 2.1415e-05, 2.9953e-05], device='cuda:1') +2023-03-20 17:55:20,406 INFO [train.py:901] (1/2) Epoch 2, batch 1600, loss[loss=0.3358, simple_loss=0.3556, pruned_loss=0.158, over 7361.00 frames. ], tot_loss[loss=0.298, simple_loss=0.3338, pruned_loss=0.131, over 1438547.10 frames. ], batch size: 73, lr: 4.30e-02, grad_scale: 8.0 +2023-03-20 17:55:27,449 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 17:55:28,442 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 17:55:31,498 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 17:55:31,644 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4407, 1.7432, 2.2316, 1.9637, 1.6941, 1.7353, 2.0964, 1.6930], + device='cuda:1'), covar=tensor([0.0755, 0.0630, 0.0255, 0.0327, 0.0715, 0.0855, 0.0250, 0.0467], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0036, 0.0042, 0.0043, 0.0040, 0.0042, 0.0037, 0.0037], + device='cuda:1'), out_proj_covar=tensor([4.0673e-05, 3.7537e-05, 4.8782e-05, 4.2553e-05, 4.1988e-05, 4.2169e-05, + 3.4337e-05, 3.2091e-05], device='cuda:1') +2023-03-20 17:55:41,793 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 17:55:46,531 INFO [train.py:901] (1/2) Epoch 2, batch 1650, loss[loss=0.2922, simple_loss=0.3384, pruned_loss=0.123, over 7308.00 frames. ], tot_loss[loss=0.2971, simple_loss=0.3333, pruned_loss=0.1304, over 1440586.24 frames. ], batch size: 80, lr: 4.29e-02, grad_scale: 8.0 +2023-03-20 17:55:46,544 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 17:55:55,367 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 17:55:58,496 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4497.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:56:01,467 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.052e+02 3.748e+02 4.507e+02 5.653e+02 9.298e+02, threshold=9.013e+02, percent-clipped=1.0 +2023-03-20 17:56:12,924 INFO [train.py:901] (1/2) Epoch 2, batch 1700, loss[loss=0.2991, simple_loss=0.3408, pruned_loss=0.1287, over 7277.00 frames. ], tot_loss[loss=0.2966, simple_loss=0.3333, pruned_loss=0.13, over 1441513.07 frames. ], batch size: 77, lr: 4.28e-02, grad_scale: 8.0 +2023-03-20 17:56:12,927 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 17:56:16,985 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 17:56:27,926 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 17:56:39,097 INFO [train.py:901] (1/2) Epoch 2, batch 1750, loss[loss=0.3067, simple_loss=0.3455, pruned_loss=0.134, over 7337.00 frames. ], tot_loss[loss=0.2959, simple_loss=0.3328, pruned_loss=0.1295, over 1443741.91 frames. ], batch size: 54, lr: 4.27e-02, grad_scale: 8.0 +2023-03-20 17:56:43,365 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4582.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:56:54,027 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 17:56:54,507 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.543e+02 3.909e+02 5.009e+02 6.432e+02 1.209e+03, threshold=1.002e+03, percent-clipped=8.0 +2023-03-20 17:56:55,048 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 17:57:02,427 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8404, 2.6061, 1.9982, 2.7447, 2.1133, 2.3377, 2.6693, 2.4331], + device='cuda:1'), covar=tensor([0.0670, 0.0392, 0.0689, 0.0378, 0.0757, 0.0494, 0.0332, 0.0478], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0026, 0.0030, 0.0030, 0.0026, 0.0024, 0.0028, 0.0025], + device='cuda:1'), out_proj_covar=tensor([3.0812e-05, 2.6930e-05, 3.1574e-05, 3.2231e-05, 2.9233e-05, 2.9301e-05, + 2.9084e-05, 3.0380e-05], device='cuda:1') +2023-03-20 17:57:05,835 INFO [train.py:901] (1/2) Epoch 2, batch 1800, loss[loss=0.2271, simple_loss=0.2591, pruned_loss=0.09756, over 7025.00 frames. ], tot_loss[loss=0.2939, simple_loss=0.3312, pruned_loss=0.1283, over 1444179.66 frames. ], batch size: 35, lr: 4.25e-02, grad_scale: 8.0 +2023-03-20 17:57:08,431 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=4630.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:57:08,794 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.16 vs. limit=5.0 +2023-03-20 17:57:15,442 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.05 vs. limit=2.0 +2023-03-20 17:57:16,202 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4644.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:57:16,651 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 17:57:17,330 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8863, 1.8845, 1.2456, 1.3981, 1.8600, 1.3544, 1.3479, 2.0747], + device='cuda:1'), covar=tensor([0.0234, 0.0203, 0.0613, 0.0332, 0.0220, 0.0469, 0.0386, 0.0142], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0027, 0.0030, 0.0030, 0.0030, 0.0030, 0.0030, 0.0023], + device='cuda:1'), out_proj_covar=tensor([2.5416e-05, 2.2413e-05, 2.7309e-05, 2.6481e-05, 2.6099e-05, 2.7302e-05, + 2.7114e-05, 1.9476e-05], device='cuda:1') +2023-03-20 17:57:24,292 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 17:57:31,219 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 17:57:32,697 INFO [train.py:901] (1/2) Epoch 2, batch 1850, loss[loss=0.2275, simple_loss=0.2611, pruned_loss=0.097, over 6970.00 frames. ], tot_loss[loss=0.2927, simple_loss=0.3304, pruned_loss=0.1275, over 1444630.82 frames. ], batch size: 35, lr: 4.24e-02, grad_scale: 8.0 +2023-03-20 17:57:40,910 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 17:57:47,106 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.532e+02 3.943e+02 4.713e+02 6.122e+02 1.050e+03, threshold=9.426e+02, percent-clipped=1.0 +2023-03-20 17:57:53,582 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.17 vs. limit=2.0 +2023-03-20 17:57:56,381 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 17:57:58,030 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3074, 4.6995, 4.5842, 4.9014, 4.4365, 4.1211, 4.8183, 4.7813], + device='cuda:1'), covar=tensor([0.0377, 0.0196, 0.0331, 0.0345, 0.0624, 0.0375, 0.0345, 0.0302], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0047, 0.0053, 0.0049, 0.0050, 0.0053, 0.0048, 0.0049], + device='cuda:1'), out_proj_covar=tensor([6.9651e-05, 5.3053e-05, 6.4139e-05, 6.3504e-05, 6.0698e-05, 6.5817e-05, + 5.8078e-05, 6.1246e-05], device='cuda:1') +2023-03-20 17:57:58,958 INFO [train.py:901] (1/2) Epoch 2, batch 1900, loss[loss=0.3203, simple_loss=0.3503, pruned_loss=0.1451, over 7230.00 frames. ], tot_loss[loss=0.2924, simple_loss=0.3304, pruned_loss=0.1273, over 1444915.40 frames. ], batch size: 93, lr: 4.23e-02, grad_scale: 8.0 +2023-03-20 17:58:20,487 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 17:58:23,203 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 17:58:25,288 INFO [train.py:901] (1/2) Epoch 2, batch 1950, loss[loss=0.2672, simple_loss=0.3149, pruned_loss=0.1097, over 7337.00 frames. ], tot_loss[loss=0.2908, simple_loss=0.3286, pruned_loss=0.1265, over 1442292.92 frames. ], batch size: 54, lr: 4.22e-02, grad_scale: 8.0 +2023-03-20 17:58:29,883 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3578, 4.5170, 4.5272, 4.7445, 4.2587, 3.9643, 4.7072, 4.6176], + device='cuda:1'), covar=tensor([0.0369, 0.0213, 0.0280, 0.0350, 0.0570, 0.0394, 0.0310, 0.0288], + device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0049, 0.0054, 0.0051, 0.0049, 0.0054, 0.0049, 0.0050], + device='cuda:1'), out_proj_covar=tensor([7.2649e-05, 5.5842e-05, 6.4933e-05, 6.5481e-05, 6.1303e-05, 6.6594e-05, + 5.9251e-05, 6.3203e-05], device='cuda:1') +2023-03-20 17:58:34,346 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 17:58:36,493 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4797.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:58:38,792 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 17:58:39,750 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.425e+02 3.763e+02 4.603e+02 6.493e+02 1.635e+03, threshold=9.206e+02, percent-clipped=4.0 +2023-03-20 17:58:39,792 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 17:58:51,641 INFO [train.py:901] (1/2) Epoch 2, batch 2000, loss[loss=0.2837, simple_loss=0.3111, pruned_loss=0.1281, over 7222.00 frames. ], tot_loss[loss=0.2909, simple_loss=0.3288, pruned_loss=0.1265, over 1442926.28 frames. ], batch size: 45, lr: 4.21e-02, grad_scale: 8.0 +2023-03-20 17:58:58,063 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 17:59:02,685 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=4845.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 17:59:09,319 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 17:59:13,002 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6247, 3.5038, 3.4944, 3.3448, 3.2335, 3.4283, 1.5528, 3.8862], + device='cuda:1'), covar=tensor([0.0046, 0.0196, 0.0085, 0.0107, 0.0053, 0.0101, 0.1360, 0.0082], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0026, 0.0026, 0.0024, 0.0024, 0.0027, 0.0049, 0.0027], + device='cuda:1'), out_proj_covar=tensor([2.3783e-05, 3.2155e-05, 2.7391e-05, 2.6857e-05, 2.2615e-05, 2.9152e-05, + 6.9865e-05, 3.1306e-05], device='cuda:1') +2023-03-20 17:59:17,045 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6317, 4.6776, 4.7529, 4.5367, 4.4960, 4.7585, 4.8016, 4.9745], + device='cuda:1'), covar=tensor([0.0238, 0.0198, 0.0139, 0.0220, 0.0268, 0.0135, 0.0369, 0.0209], + device='cuda:1'), in_proj_covar=tensor([0.0064, 0.0068, 0.0061, 0.0084, 0.0069, 0.0057, 0.0056, 0.0056], + device='cuda:1'), out_proj_covar=tensor([8.5036e-05, 8.8953e-05, 7.9536e-05, 1.1790e-04, 9.3509e-05, 7.5479e-05, + 7.2149e-05, 7.3531e-05], device='cuda:1') +2023-03-20 17:59:17,524 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 17:59:18,002 INFO [train.py:901] (1/2) Epoch 2, batch 2050, loss[loss=0.2677, simple_loss=0.3163, pruned_loss=0.1095, over 7371.00 frames. ], tot_loss[loss=0.2893, simple_loss=0.328, pruned_loss=0.1253, over 1441921.03 frames. ], batch size: 65, lr: 4.20e-02, grad_scale: 8.0 +2023-03-20 17:59:32,997 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.296e+02 3.547e+02 4.263e+02 5.510e+02 1.038e+03, threshold=8.526e+02, percent-clipped=1.0 +2023-03-20 17:59:44,249 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 +2023-03-20 17:59:44,894 INFO [train.py:901] (1/2) Epoch 2, batch 2100, loss[loss=0.2563, simple_loss=0.2908, pruned_loss=0.1109, over 7142.00 frames. ], tot_loss[loss=0.2897, simple_loss=0.3286, pruned_loss=0.1254, over 1440924.92 frames. ], batch size: 39, lr: 4.19e-02, grad_scale: 8.0 +2023-03-20 17:59:45,539 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2965, 3.1824, 2.7378, 3.7971, 2.7266, 3.5708, 3.5212, 3.5954], + device='cuda:1'), covar=tensor([0.0197, 0.0481, 0.3214, 0.0136, 0.4546, 0.0159, 0.0572, 0.0113], + device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0120, 0.0240, 0.0081, 0.0201, 0.0093, 0.0143, 0.0081], + device='cuda:1'), out_proj_covar=tensor([6.0789e-05, 9.7534e-05, 1.7321e-04, 5.6368e-05, 1.5902e-04, 6.1695e-05, + 1.0798e-04, 5.9973e-05], device='cuda:1') +2023-03-20 17:59:47,700 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-03-20 17:59:48,582 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6058, 3.7445, 3.1845, 3.9070, 3.2964, 3.8691, 4.2647, 4.0039], + device='cuda:1'), covar=tensor([0.0246, 0.0528, 0.2937, 0.0154, 0.3966, 0.0127, 0.0244, 0.0102], + device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0118, 0.0238, 0.0081, 0.0200, 0.0092, 0.0142, 0.0080], + device='cuda:1'), out_proj_covar=tensor([6.0617e-05, 9.6534e-05, 1.7212e-04, 5.6400e-05, 1.5809e-04, 6.1399e-05, + 1.0731e-04, 5.9923e-05], device='cuda:1') +2023-03-20 17:59:51,541 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 17:59:54,111 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 17:59:54,711 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4944.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:00:10,565 INFO [train.py:901] (1/2) Epoch 2, batch 2150, loss[loss=0.2865, simple_loss=0.3273, pruned_loss=0.1228, over 7346.00 frames. ], tot_loss[loss=0.2911, simple_loss=0.3297, pruned_loss=0.1262, over 1441082.01 frames. ], batch size: 63, lr: 4.18e-02, grad_scale: 8.0 +2023-03-20 18:00:19,867 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=4992.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:00:24,967 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1219, 2.7330, 3.0686, 3.0259, 2.2538, 3.1657, 3.0084, 2.8109], + device='cuda:1'), covar=tensor([0.1626, 0.0731, 0.1892, 0.0155, 0.0171, 0.0124, 0.0090, 0.0171], + device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0080, 0.0122, 0.0048, 0.0052, 0.0060, 0.0047, 0.0050], + device='cuda:1'), out_proj_covar=tensor([8.6237e-05, 5.8399e-05, 9.9068e-05, 3.6590e-05, 3.6532e-05, 4.2886e-05, + 3.3389e-05, 3.4925e-05], device='cuda:1') +2023-03-20 18:00:25,812 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.955e+02 4.119e+02 4.986e+02 6.315e+02 1.234e+03, threshold=9.973e+02, percent-clipped=7.0 +2023-03-20 18:00:31,914 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 +2023-03-20 18:00:33,642 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9362, 4.3007, 4.2366, 4.3588, 4.1032, 3.7480, 4.4660, 4.3078], + device='cuda:1'), covar=tensor([0.0505, 0.0233, 0.0324, 0.0547, 0.0457, 0.0492, 0.0293, 0.0296], + device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0049, 0.0056, 0.0053, 0.0053, 0.0058, 0.0051, 0.0051], + device='cuda:1'), out_proj_covar=tensor([7.7650e-05, 5.6303e-05, 6.7370e-05, 7.1312e-05, 6.7598e-05, 7.2825e-05, + 6.0884e-05, 6.6094e-05], device='cuda:1') +2023-03-20 18:00:37,642 INFO [train.py:901] (1/2) Epoch 2, batch 2200, loss[loss=0.2792, simple_loss=0.3196, pruned_loss=0.1194, over 7262.00 frames. ], tot_loss[loss=0.291, simple_loss=0.3298, pruned_loss=0.1261, over 1443383.41 frames. ], batch size: 47, lr: 4.17e-02, grad_scale: 8.0 +2023-03-20 18:00:40,189 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 18:00:53,898 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 +2023-03-20 18:01:03,873 INFO [train.py:901] (1/2) Epoch 2, batch 2250, loss[loss=0.2611, simple_loss=0.2756, pruned_loss=0.1233, over 6249.00 frames. ], tot_loss[loss=0.2911, simple_loss=0.3297, pruned_loss=0.1263, over 1443140.22 frames. ], batch size: 27, lr: 4.16e-02, grad_scale: 8.0 +2023-03-20 18:01:05,017 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0155, 3.7534, 2.9788, 4.1782, 3.3876, 4.0837, 4.3042, 4.1361], + device='cuda:1'), covar=tensor([0.0147, 0.0472, 0.3097, 0.0126, 0.3337, 0.0109, 0.0174, 0.0128], + device='cuda:1'), in_proj_covar=tensor([0.0089, 0.0124, 0.0243, 0.0083, 0.0206, 0.0093, 0.0145, 0.0082], + device='cuda:1'), out_proj_covar=tensor([6.1589e-05, 9.9856e-05, 1.7591e-04, 5.8739e-05, 1.6197e-04, 6.2700e-05, + 1.0981e-04, 5.9761e-05], device='cuda:1') +2023-03-20 18:01:12,188 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2344, 3.2853, 3.4448, 3.4413, 3.2635, 3.3630, 3.6391, 3.1598], + device='cuda:1'), covar=tensor([0.0104, 0.0143, 0.0135, 0.0103, 0.0148, 0.0131, 0.0148, 0.0144], + device='cuda:1'), in_proj_covar=tensor([0.0016, 0.0017, 0.0017, 0.0017, 0.0016, 0.0017, 0.0021, 0.0019], + device='cuda:1'), out_proj_covar=tensor([2.1691e-05, 2.9689e-05, 2.4534e-05, 2.8189e-05, 2.5760e-05, 2.6386e-05, + 3.3002e-05, 2.9753e-05], device='cuda:1') +2023-03-20 18:01:14,259 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4457, 3.8219, 3.6023, 3.7893, 3.4897, 3.8096, 4.1868, 4.2209], + device='cuda:1'), covar=tensor([0.0371, 0.0271, 0.0397, 0.0309, 0.0463, 0.0366, 0.0303, 0.0257], + device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0063, 0.0061, 0.0079, 0.0065, 0.0055, 0.0050, 0.0053], + device='cuda:1'), out_proj_covar=tensor([8.0406e-05, 8.3149e-05, 7.9859e-05, 1.1383e-04, 8.9157e-05, 7.3279e-05, + 6.6695e-05, 7.1408e-05], device='cuda:1') +2023-03-20 18:01:14,670 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 18:01:14,681 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 18:01:18,730 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.254e+02 3.930e+02 5.031e+02 6.019e+02 2.322e+03, threshold=1.006e+03, percent-clipped=4.0 +2023-03-20 18:01:19,337 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.8210, 4.9536, 5.0040, 5.2104, 4.6936, 4.5678, 5.2495, 5.0188], + device='cuda:1'), covar=tensor([0.0327, 0.0210, 0.0223, 0.0479, 0.0382, 0.0333, 0.0231, 0.0260], + device='cuda:1'), in_proj_covar=tensor([0.0064, 0.0051, 0.0057, 0.0055, 0.0052, 0.0059, 0.0053, 0.0053], + device='cuda:1'), out_proj_covar=tensor([8.0886e-05, 5.9694e-05, 7.0734e-05, 7.4565e-05, 6.7969e-05, 7.5354e-05, + 6.4202e-05, 6.8692e-05], device='cuda:1') +2023-03-20 18:01:28,441 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 18:01:29,957 INFO [train.py:901] (1/2) Epoch 2, batch 2300, loss[loss=0.2946, simple_loss=0.3325, pruned_loss=0.1283, over 7281.00 frames. ], tot_loss[loss=0.2904, simple_loss=0.3293, pruned_loss=0.1257, over 1442630.37 frames. ], batch size: 66, lr: 4.15e-02, grad_scale: 8.0 +2023-03-20 18:01:32,900 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.17 vs. limit=2.0 +2023-03-20 18:01:35,945 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 +2023-03-20 18:01:39,009 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 +2023-03-20 18:01:48,409 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4614, 3.8269, 3.7953, 3.6138, 3.6365, 3.7428, 4.1517, 4.2458], + device='cuda:1'), covar=tensor([0.0421, 0.0307, 0.0271, 0.0334, 0.0356, 0.0414, 0.0311, 0.0214], + device='cuda:1'), in_proj_covar=tensor([0.0063, 0.0066, 0.0063, 0.0081, 0.0066, 0.0057, 0.0054, 0.0054], + device='cuda:1'), out_proj_covar=tensor([8.3767e-05, 8.6984e-05, 8.4231e-05, 1.1950e-04, 9.1277e-05, 7.5826e-05, + 7.2429e-05, 7.3297e-05], device='cuda:1') +2023-03-20 18:01:53,610 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2701, 2.3201, 2.8186, 2.7124, 2.2655, 3.2633, 2.5589, 2.8620], + device='cuda:1'), covar=tensor([0.1805, 0.0923, 0.2501, 0.0250, 0.0139, 0.0180, 0.0058, 0.0082], + device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0081, 0.0124, 0.0049, 0.0052, 0.0062, 0.0047, 0.0050], + device='cuda:1'), out_proj_covar=tensor([8.7312e-05, 6.0316e-05, 1.0080e-04, 3.7016e-05, 3.8005e-05, 4.4843e-05, + 3.2475e-05, 3.4910e-05], device='cuda:1') +2023-03-20 18:01:55,974 INFO [train.py:901] (1/2) Epoch 2, batch 2350, loss[loss=0.2766, simple_loss=0.3168, pruned_loss=0.1182, over 7318.00 frames. ], tot_loss[loss=0.2901, simple_loss=0.3291, pruned_loss=0.1255, over 1443199.01 frames. ], batch size: 49, lr: 4.14e-02, grad_scale: 8.0 +2023-03-20 18:02:11,288 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.216e+02 3.739e+02 4.530e+02 5.733e+02 1.414e+03, threshold=9.059e+02, percent-clipped=1.0 +2023-03-20 18:02:15,347 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 18:02:18,504 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5935, 4.8303, 4.8928, 5.0376, 4.4744, 4.3061, 5.0323, 4.8094], + device='cuda:1'), covar=tensor([0.0397, 0.0209, 0.0230, 0.0388, 0.0597, 0.0347, 0.0286, 0.0361], + device='cuda:1'), in_proj_covar=tensor([0.0064, 0.0050, 0.0057, 0.0054, 0.0052, 0.0059, 0.0053, 0.0052], + device='cuda:1'), out_proj_covar=tensor([8.2545e-05, 5.9928e-05, 7.1240e-05, 7.3412e-05, 6.7477e-05, 7.6119e-05, + 6.6192e-05, 6.7474e-05], device='cuda:1') +2023-03-20 18:02:22,461 INFO [train.py:901] (1/2) Epoch 2, batch 2400, loss[loss=0.2732, simple_loss=0.3263, pruned_loss=0.1101, over 7304.00 frames. ], tot_loss[loss=0.2898, simple_loss=0.3289, pruned_loss=0.1253, over 1442933.66 frames. ], batch size: 59, lr: 4.13e-02, grad_scale: 8.0 +2023-03-20 18:02:22,463 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 18:02:23,766 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.56 vs. limit=5.0 +2023-03-20 18:02:34,093 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0646, 3.4001, 3.2954, 3.5139, 3.8296, 3.3859, 2.7726, 3.4850], + device='cuda:1'), covar=tensor([0.0986, 0.0329, 0.2318, 0.0111, 0.0063, 0.0140, 0.0121, 0.0090], + device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0086, 0.0134, 0.0052, 0.0052, 0.0063, 0.0050, 0.0054], + device='cuda:1'), out_proj_covar=tensor([9.1428e-05, 6.4051e-05, 1.0844e-04, 4.0509e-05, 3.8750e-05, 4.6085e-05, + 3.5058e-05, 3.7939e-05], device='cuda:1') +2023-03-20 18:02:34,438 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 18:02:37,587 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 18:02:49,398 INFO [train.py:901] (1/2) Epoch 2, batch 2450, loss[loss=0.2732, simple_loss=0.3164, pruned_loss=0.115, over 7299.00 frames. ], tot_loss[loss=0.2867, simple_loss=0.3266, pruned_loss=0.1234, over 1442521.36 frames. ], batch size: 86, lr: 4.12e-02, grad_scale: 8.0 +2023-03-20 18:03:03,829 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.303e+02 3.507e+02 4.658e+02 5.816e+02 1.903e+03, threshold=9.317e+02, percent-clipped=8.0 +2023-03-20 18:03:05,439 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 18:03:15,080 INFO [train.py:901] (1/2) Epoch 2, batch 2500, loss[loss=0.2796, simple_loss=0.3194, pruned_loss=0.1199, over 7306.00 frames. ], tot_loss[loss=0.2854, simple_loss=0.3252, pruned_loss=0.1228, over 1443823.73 frames. ], batch size: 68, lr: 4.11e-02, grad_scale: 8.0 +2023-03-20 18:03:31,871 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 18:03:41,955 INFO [train.py:901] (1/2) Epoch 2, batch 2550, loss[loss=0.2448, simple_loss=0.297, pruned_loss=0.09634, over 7287.00 frames. ], tot_loss[loss=0.2852, simple_loss=0.3252, pruned_loss=0.1226, over 1444254.45 frames. ], batch size: 68, lr: 4.10e-02, grad_scale: 8.0 +2023-03-20 18:03:56,653 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 4.460e+02 5.219e+02 6.582e+02 1.378e+03, threshold=1.044e+03, percent-clipped=5.0 +2023-03-20 18:04:08,624 INFO [train.py:901] (1/2) Epoch 2, batch 2600, loss[loss=0.2556, simple_loss=0.3067, pruned_loss=0.1023, over 7291.00 frames. ], tot_loss[loss=0.2825, simple_loss=0.3232, pruned_loss=0.1209, over 1442789.31 frames. ], batch size: 68, lr: 4.09e-02, grad_scale: 16.0 +2023-03-20 18:04:34,255 INFO [train.py:901] (1/2) Epoch 2, batch 2650, loss[loss=0.2853, simple_loss=0.315, pruned_loss=0.1278, over 7267.00 frames. ], tot_loss[loss=0.2829, simple_loss=0.3236, pruned_loss=0.1211, over 1444365.20 frames. ], batch size: 47, lr: 4.08e-02, grad_scale: 16.0 +2023-03-20 18:04:48,315 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.928e+02 3.829e+02 4.633e+02 5.948e+02 1.485e+03, threshold=9.266e+02, percent-clipped=1.0 +2023-03-20 18:04:59,253 INFO [train.py:901] (1/2) Epoch 2, batch 2700, loss[loss=0.301, simple_loss=0.3359, pruned_loss=0.133, over 7337.00 frames. ], tot_loss[loss=0.2848, simple_loss=0.3251, pruned_loss=0.1222, over 1442922.70 frames. ], batch size: 51, lr: 4.07e-02, grad_scale: 16.0 +2023-03-20 18:05:07,568 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.10 vs. limit=2.0 +2023-03-20 18:05:09,330 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6122, 1.2794, 2.2798, 1.6714, 1.7691, 1.4376, 1.8392, 1.6883], + device='cuda:1'), covar=tensor([0.0649, 0.1992, 0.0261, 0.0570, 0.0577, 0.1324, 0.0187, 0.0331], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0038, 0.0035, 0.0041, 0.0032, 0.0040, 0.0034, 0.0034], + device='cuda:1'), out_proj_covar=tensor([3.9599e-05, 4.3663e-05, 4.6381e-05, 4.4394e-05, 3.8975e-05, 4.6335e-05, + 3.2585e-05, 3.5354e-05], device='cuda:1') +2023-03-20 18:05:24,206 INFO [train.py:901] (1/2) Epoch 2, batch 2750, loss[loss=0.2761, simple_loss=0.3229, pruned_loss=0.1146, over 7310.00 frames. ], tot_loss[loss=0.2857, simple_loss=0.3259, pruned_loss=0.1227, over 1444396.08 frames. ], batch size: 83, lr: 4.06e-02, grad_scale: 16.0 +2023-03-20 18:05:29,431 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-03-20 18:05:34,083 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 +2023-03-20 18:05:38,365 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.897e+02 3.942e+02 4.778e+02 5.507e+02 1.235e+03, threshold=9.556e+02, percent-clipped=3.0 +2023-03-20 18:05:43,845 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6549, 3.6940, 3.9026, 3.7236, 3.6616, 3.7865, 4.1153, 4.1948], + device='cuda:1'), covar=tensor([0.0384, 0.0302, 0.0260, 0.0293, 0.0356, 0.0496, 0.0333, 0.0242], + device='cuda:1'), in_proj_covar=tensor([0.0064, 0.0065, 0.0058, 0.0078, 0.0065, 0.0055, 0.0051, 0.0054], + device='cuda:1'), out_proj_covar=tensor([8.9834e-05, 8.8916e-05, 7.9991e-05, 1.1952e-04, 9.5240e-05, 7.6241e-05, + 7.1399e-05, 7.6322e-05], device='cuda:1') +2023-03-20 18:05:49,311 INFO [train.py:901] (1/2) Epoch 2, batch 2800, loss[loss=0.282, simple_loss=0.3254, pruned_loss=0.1193, over 7324.00 frames. ], tot_loss[loss=0.2861, simple_loss=0.3267, pruned_loss=0.1228, over 1446785.67 frames. ], batch size: 59, lr: 4.05e-02, grad_scale: 16.0 +2023-03-20 18:05:53,325 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9165, 3.7263, 3.5783, 3.3786, 3.6350, 3.4923, 4.0141, 3.8927], + device='cuda:1'), covar=tensor([0.0215, 0.0155, 0.0283, 0.0315, 0.0226, 0.0224, 0.0220, 0.0164], + device='cuda:1'), in_proj_covar=tensor([0.0026, 0.0028, 0.0030, 0.0028, 0.0035, 0.0033, 0.0028, 0.0026], + device='cuda:1'), out_proj_covar=tensor([4.0090e-05, 4.1752e-05, 5.2313e-05, 4.8110e-05, 5.7603e-05, 5.1379e-05, + 4.6435e-05, 3.6379e-05], device='cuda:1') +2023-03-20 18:05:54,190 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-03-20 18:06:14,502 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 18:06:14,554 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 18:06:14,578 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 18:06:14,594 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 18:06:15,074 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 18:06:15,089 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 18:06:15,100 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 18:06:15,201 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 18:06:15,276 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 18:06:15,290 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 18:06:15,307 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 18:06:15,470 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 18:06:15,477 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 18:06:15,484 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 18:06:15,620 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 18:06:15,623 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 18:06:16,009 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 18:06:23,285 INFO [train.py:901] (1/2) Epoch 3, batch 0, loss[loss=0.2506, simple_loss=0.2989, pruned_loss=0.1012, over 7140.00 frames. ], tot_loss[loss=0.2506, simple_loss=0.2989, pruned_loss=0.1012, over 7140.00 frames. ], batch size: 39, lr: 3.96e-02, grad_scale: 16.0 +2023-03-20 18:06:23,285 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 18:06:33,370 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3893, 2.4685, 2.3882, 1.9977, 1.4730, 1.7167, 2.6606, 1.7901], + device='cuda:1'), covar=tensor([0.0684, 0.0439, 0.0321, 0.0673, 0.1229, 0.0903, 0.0467, 0.0976], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0033, 0.0031, 0.0037, 0.0037, 0.0037, 0.0039, 0.0031], + device='cuda:1'), out_proj_covar=tensor([4.5547e-05, 3.9348e-05, 3.6175e-05, 4.1247e-05, 4.3620e-05, 4.3305e-05, + 4.5908e-05, 3.7488e-05], device='cuda:1') +2023-03-20 18:06:48,491 INFO [train.py:935] (1/2) Epoch 3, validation: loss=0.2204, simple_loss=0.3002, pruned_loss=0.07031, over 1622729.00 frames. +2023-03-20 18:06:48,491 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-20 18:06:49,113 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5650.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 18:06:55,579 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 18:07:00,482 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4458, 2.6780, 2.2614, 2.3546, 1.8505, 1.7314, 2.8328, 2.2289], + device='cuda:1'), covar=tensor([0.0903, 0.0587, 0.1144, 0.0639, 0.1020, 0.0908, 0.0504, 0.0669], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0034, 0.0032, 0.0038, 0.0036, 0.0038, 0.0040, 0.0031], + device='cuda:1'), out_proj_covar=tensor([4.7260e-05, 4.0082e-05, 3.7468e-05, 4.2416e-05, 4.2752e-05, 4.4405e-05, + 4.7436e-05, 3.7577e-05], device='cuda:1') +2023-03-20 18:07:07,897 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 18:07:15,077 INFO [train.py:901] (1/2) Epoch 3, batch 50, loss[loss=0.2848, simple_loss=0.3274, pruned_loss=0.1211, over 7294.00 frames. ], tot_loss[loss=0.2819, simple_loss=0.3227, pruned_loss=0.1205, over 324765.99 frames. ], batch size: 66, lr: 3.95e-02, grad_scale: 16.0 +2023-03-20 18:07:16,074 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 18:07:17,059 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.426e+02 3.989e+02 5.085e+02 6.289e+02 1.206e+03, threshold=1.017e+03, percent-clipped=2.0 +2023-03-20 18:07:18,595 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 18:07:21,273 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5711.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 18:07:22,126 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 18:07:38,376 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 18:07:38,388 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 18:07:38,716 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 18:07:40,414 INFO [train.py:901] (1/2) Epoch 3, batch 100, loss[loss=0.3554, simple_loss=0.3743, pruned_loss=0.1682, over 6867.00 frames. ], tot_loss[loss=0.2768, simple_loss=0.3181, pruned_loss=0.1178, over 570709.43 frames. ], batch size: 107, lr: 3.95e-02, grad_scale: 16.0 +2023-03-20 18:07:43,589 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5755.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:07:54,654 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 +2023-03-20 18:08:07,074 INFO [train.py:901] (1/2) Epoch 3, batch 150, loss[loss=0.3496, simple_loss=0.3789, pruned_loss=0.1602, over 6702.00 frames. ], tot_loss[loss=0.2783, simple_loss=0.32, pruned_loss=0.1183, over 765152.80 frames. ], batch size: 106, lr: 3.94e-02, grad_scale: 16.0 +2023-03-20 18:08:09,472 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.257e+02 3.860e+02 4.707e+02 5.554e+02 1.675e+03, threshold=9.414e+02, percent-clipped=1.0 +2023-03-20 18:08:16,299 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5816.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:08:33,339 INFO [train.py:901] (1/2) Epoch 3, batch 200, loss[loss=0.2759, simple_loss=0.3271, pruned_loss=0.1124, over 7322.00 frames. ], tot_loss[loss=0.2821, simple_loss=0.3228, pruned_loss=0.1207, over 917364.29 frames. ], batch size: 83, lr: 3.93e-02, grad_scale: 16.0 +2023-03-20 18:08:37,942 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 18:08:38,049 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5858.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:08:39,021 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7936, 3.8844, 3.9472, 3.3849, 3.1895, 3.3884, 1.5580, 4.0996], + device='cuda:1'), covar=tensor([0.0065, 0.0219, 0.0078, 0.0121, 0.0118, 0.0165, 0.1479, 0.0072], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0029, 0.0031, 0.0029, 0.0030, 0.0033, 0.0060, 0.0032], + device='cuda:1'), out_proj_covar=tensor([2.9142e-05, 3.9079e-05, 3.5171e-05, 3.7232e-05, 3.0564e-05, 3.8696e-05, + 8.4808e-05, 3.8599e-05], device='cuda:1') +2023-03-20 18:08:39,665 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 +2023-03-20 18:08:43,622 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 18:08:48,420 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.55 vs. limit=2.0 +2023-03-20 18:08:50,073 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 18:08:59,359 INFO [train.py:901] (1/2) Epoch 3, batch 250, loss[loss=0.2571, simple_loss=0.3077, pruned_loss=0.1033, over 7269.00 frames. ], tot_loss[loss=0.2815, simple_loss=0.3228, pruned_loss=0.1201, over 1036327.32 frames. ], batch size: 57, lr: 3.92e-02, grad_scale: 16.0 +2023-03-20 18:09:01,493 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.164e+02 3.767e+02 5.039e+02 6.362e+02 1.347e+03, threshold=1.008e+03, percent-clipped=5.0 +2023-03-20 18:09:03,640 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 18:09:05,307 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3138, 2.5290, 1.8437, 1.7393, 2.5119, 2.3459, 1.3838, 2.4262], + device='cuda:1'), covar=tensor([0.0566, 0.0144, 0.0991, 0.0956, 0.0293, 0.0320, 0.1042, 0.0447], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0024, 0.0025, 0.0027, 0.0026, 0.0025, 0.0028, 0.0031], + device='cuda:1'), out_proj_covar=tensor([3.3985e-05, 2.7085e-05, 3.3900e-05, 3.5047e-05, 3.3553e-05, 3.3526e-05, + 3.4285e-05, 4.1376e-05], device='cuda:1') +2023-03-20 18:09:06,069 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.02 vs. limit=2.0 +2023-03-20 18:09:09,915 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5919.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:09:25,528 INFO [train.py:901] (1/2) Epoch 3, batch 300, loss[loss=0.2685, simple_loss=0.3198, pruned_loss=0.1086, over 7320.00 frames. ], tot_loss[loss=0.2784, simple_loss=0.3202, pruned_loss=0.1183, over 1125301.27 frames. ], batch size: 75, lr: 3.91e-02, grad_scale: 16.0 +2023-03-20 18:09:25,545 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 18:09:35,785 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=6.26 vs. limit=5.0 +2023-03-20 18:09:36,465 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 18:09:45,594 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8474, 2.5660, 2.5878, 2.4191, 1.0313, 2.2351, 2.9721, 1.3843], + device='cuda:1'), covar=tensor([0.0812, 0.0612, 0.0394, 0.0370, 0.2024, 0.0722, 0.0511, 0.2334], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0034, 0.0031, 0.0037, 0.0034, 0.0037, 0.0039, 0.0033], + device='cuda:1'), out_proj_covar=tensor([4.8425e-05, 4.1412e-05, 3.7506e-05, 4.2316e-05, 4.2701e-05, 4.4944e-05, + 5.0151e-05, 4.1391e-05], device='cuda:1') +2023-03-20 18:09:47,132 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5990.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:09:51,491 INFO [train.py:901] (1/2) Epoch 3, batch 350, loss[loss=0.2826, simple_loss=0.33, pruned_loss=0.1176, over 7244.00 frames. ], tot_loss[loss=0.279, simple_loss=0.3207, pruned_loss=0.1186, over 1196356.17 frames. ], batch size: 55, lr: 3.90e-02, grad_scale: 16.0 +2023-03-20 18:09:53,928 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.502e+02 3.944e+02 4.876e+02 5.780e+02 1.019e+03, threshold=9.752e+02, percent-clipped=1.0 +2023-03-20 18:09:55,520 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6006.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 18:10:11,573 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 18:10:14,863 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6043.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:10:17,775 INFO [train.py:901] (1/2) Epoch 3, batch 400, loss[loss=0.3056, simple_loss=0.3434, pruned_loss=0.1339, over 7222.00 frames. ], tot_loss[loss=0.278, simple_loss=0.3199, pruned_loss=0.118, over 1252024.91 frames. ], batch size: 93, lr: 3.89e-02, grad_scale: 16.0 +2023-03-20 18:10:19,594 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6051.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:10:43,822 INFO [train.py:901] (1/2) Epoch 3, batch 450, loss[loss=0.2812, simple_loss=0.326, pruned_loss=0.1182, over 7255.00 frames. ], tot_loss[loss=0.2782, simple_loss=0.3199, pruned_loss=0.1182, over 1295023.25 frames. ], batch size: 89, lr: 3.88e-02, grad_scale: 16.0 +2023-03-20 18:10:45,144 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.12 vs. limit=2.0 +2023-03-20 18:10:45,831 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.319e+02 3.819e+02 4.941e+02 5.954e+02 1.043e+03, threshold=9.881e+02, percent-clipped=2.0 +2023-03-20 18:10:46,511 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6104.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:10:50,581 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6111.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:10:51,138 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5454, 3.0559, 2.4879, 2.9803, 1.9881, 2.0947, 2.7228, 1.9429], + device='cuda:1'), covar=tensor([0.0564, 0.0167, 0.0310, 0.0201, 0.1123, 0.0712, 0.0318, 0.1080], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0026, 0.0029, 0.0027, 0.0025, 0.0025, 0.0029, 0.0025], + device='cuda:1'), out_proj_covar=tensor([3.7178e-05, 3.0117e-05, 3.7445e-05, 3.4989e-05, 3.3782e-05, 3.6369e-05, + 3.6300e-05, 3.7883e-05], device='cuda:1') +2023-03-20 18:10:52,055 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 18:10:52,557 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 18:11:03,550 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6135.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:11:10,619 INFO [train.py:901] (1/2) Epoch 3, batch 500, loss[loss=0.2059, simple_loss=0.2342, pruned_loss=0.08881, over 5744.00 frames. ], tot_loss[loss=0.2767, simple_loss=0.3189, pruned_loss=0.1172, over 1326610.74 frames. ], batch size: 25, lr: 3.87e-02, grad_scale: 16.0 +2023-03-20 18:11:10,961 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-03-20 18:11:17,430 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3319, 3.5654, 3.3636, 3.5658, 3.2354, 3.6755, 3.3910, 3.0013], + device='cuda:1'), covar=tensor([0.0125, 0.0193, 0.0202, 0.0197, 0.0214, 0.0137, 0.0153, 0.0259], + device='cuda:1'), in_proj_covar=tensor([0.0018, 0.0020, 0.0020, 0.0020, 0.0018, 0.0020, 0.0023, 0.0021], + device='cuda:1'), out_proj_covar=tensor([3.0551e-05, 3.9563e-05, 3.7432e-05, 3.9249e-05, 3.5200e-05, 3.4832e-05, + 4.3543e-05, 4.0938e-05], device='cuda:1') +2023-03-20 18:11:18,171 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.00 vs. limit=2.0 +2023-03-20 18:11:21,454 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6170.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:11:25,855 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 18:11:27,373 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 18:11:27,887 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 18:11:30,435 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 18:11:30,577 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0808, 2.9690, 2.5352, 2.9405, 3.0028, 2.6602, 2.6014, 3.0969], + device='cuda:1'), covar=tensor([0.0367, 0.0125, 0.0440, 0.0485, 0.0349, 0.0466, 0.0649, 0.0498], + device='cuda:1'), in_proj_covar=tensor([0.0026, 0.0023, 0.0025, 0.0025, 0.0026, 0.0026, 0.0028, 0.0028], + device='cuda:1'), out_proj_covar=tensor([3.2885e-05, 2.8498e-05, 3.4250e-05, 3.4378e-05, 3.4918e-05, 3.5663e-05, + 3.5256e-05, 3.7995e-05], device='cuda:1') +2023-03-20 18:11:33,414 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 +2023-03-20 18:11:35,174 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6196.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:11:35,538 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 18:11:36,536 INFO [train.py:901] (1/2) Epoch 3, batch 550, loss[loss=0.2883, simple_loss=0.331, pruned_loss=0.1228, over 7307.00 frames. ], tot_loss[loss=0.2767, simple_loss=0.3186, pruned_loss=0.1174, over 1353304.30 frames. ], batch size: 86, lr: 3.86e-02, grad_scale: 16.0 +2023-03-20 18:11:38,853 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.922e+02 3.999e+02 4.896e+02 5.931e+02 1.258e+03, threshold=9.791e+02, percent-clipped=1.0 +2023-03-20 18:11:39,038 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6203.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:11:43,080 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6211.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:11:44,547 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6214.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:11:47,093 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 18:11:53,924 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6231.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:11:55,862 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 18:11:59,459 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 18:12:03,035 INFO [train.py:901] (1/2) Epoch 3, batch 600, loss[loss=0.2751, simple_loss=0.3264, pruned_loss=0.1119, over 7282.00 frames. ], tot_loss[loss=0.2759, simple_loss=0.3186, pruned_loss=0.1166, over 1375971.98 frames. ], batch size: 77, lr: 3.85e-02, grad_scale: 16.0 +2023-03-20 18:12:06,028 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 +2023-03-20 18:12:06,190 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 18:12:10,968 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6264.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:12:11,937 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6266.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:12:15,019 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6272.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:12:20,260 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.17 vs. limit=2.0 +2023-03-20 18:12:22,302 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.15 vs. limit=2.0 +2023-03-20 18:12:22,503 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 18:12:27,200 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2534, 2.7814, 2.3127, 3.6409, 1.6357, 2.4728, 3.2952, 2.4009], + device='cuda:1'), covar=tensor([0.0547, 0.0440, 0.0467, 0.0087, 0.1258, 0.0848, 0.0373, 0.0709], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0026, 0.0030, 0.0025, 0.0027, 0.0024, 0.0027, 0.0026], + device='cuda:1'), out_proj_covar=tensor([3.8347e-05, 3.2548e-05, 3.9892e-05, 3.3859e-05, 3.8147e-05, 3.5192e-05, + 3.5036e-05, 3.9433e-05], device='cuda:1') +2023-03-20 18:12:29,107 INFO [train.py:901] (1/2) Epoch 3, batch 650, loss[loss=0.2627, simple_loss=0.3102, pruned_loss=0.1076, over 7323.00 frames. ], tot_loss[loss=0.2769, simple_loss=0.3195, pruned_loss=0.1171, over 1391123.47 frames. ], batch size: 49, lr: 3.84e-02, grad_scale: 16.0 +2023-03-20 18:12:31,140 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.321e+02 3.541e+02 4.453e+02 5.741e+02 1.629e+03, threshold=8.906e+02, percent-clipped=5.0 +2023-03-20 18:12:31,695 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 18:12:32,772 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6306.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 18:12:44,099 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6327.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:12:45,810 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-03-20 18:12:46,673 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2663, 2.3316, 2.8727, 2.9201, 2.8662, 2.8566, 2.1112, 2.7366], + device='cuda:1'), covar=tensor([0.1315, 0.0884, 0.1444, 0.0312, 0.0134, 0.0187, 0.0094, 0.0088], + device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0104, 0.0145, 0.0067, 0.0060, 0.0064, 0.0062, 0.0060], + device='cuda:1'), out_proj_covar=tensor([1.1730e-04, 8.3982e-05, 1.1968e-04, 5.7451e-05, 4.6615e-05, 4.9609e-05, + 5.1082e-05, 4.4227e-05], device='cuda:1') +2023-03-20 18:12:48,019 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 18:12:53,588 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6346.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:12:55,021 INFO [train.py:901] (1/2) Epoch 3, batch 700, loss[loss=0.2358, simple_loss=0.2849, pruned_loss=0.09339, over 7136.00 frames. ], tot_loss[loss=0.2784, simple_loss=0.3208, pruned_loss=0.118, over 1402748.80 frames. ], batch size: 41, lr: 3.83e-02, grad_scale: 16.0 +2023-03-20 18:12:56,564 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 18:12:57,625 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=6354.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 18:13:20,812 INFO [train.py:901] (1/2) Epoch 3, batch 750, loss[loss=0.2905, simple_loss=0.3275, pruned_loss=0.1268, over 7279.00 frames. ], tot_loss[loss=0.2795, simple_loss=0.3216, pruned_loss=0.1187, over 1411738.52 frames. ], batch size: 70, lr: 3.82e-02, grad_scale: 16.0 +2023-03-20 18:13:20,839 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 18:13:21,564 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6399.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:13:22,011 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 18:13:23,900 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.231e+02 3.739e+02 4.769e+02 6.356e+02 1.531e+03, threshold=9.539e+02, percent-clipped=9.0 +2023-03-20 18:13:27,157 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8894, 3.1531, 2.0928, 3.0105, 2.7346, 2.7229, 2.4072, 2.0078], + device='cuda:1'), covar=tensor([0.0102, 0.0487, 0.0996, 0.0226, 0.0128, 0.0136, 0.0880, 0.1021], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0055, 0.0111, 0.0058, 0.0050, 0.0052, 0.0121, 0.0108], + device='cuda:1'), out_proj_covar=tensor([4.4628e-05, 4.9992e-05, 9.7865e-05, 5.2853e-05, 4.5954e-05, 4.6566e-05, + 1.0902e-04, 9.4684e-05], device='cuda:1') +2023-03-20 18:13:28,098 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6411.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:13:31,350 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.56 vs. limit=5.0 +2023-03-20 18:13:36,756 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 18:13:37,393 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7627, 2.5212, 1.2647, 2.7332, 1.9597, 1.6460, 2.0313, 1.5611], + device='cuda:1'), covar=tensor([0.0824, 0.0247, 0.0783, 0.0164, 0.0572, 0.1012, 0.0340, 0.1018], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0026, 0.0031, 0.0025, 0.0025, 0.0025, 0.0026, 0.0025], + device='cuda:1'), out_proj_covar=tensor([3.8755e-05, 3.1980e-05, 4.2149e-05, 3.3507e-05, 3.5024e-05, 3.6579e-05, + 3.4620e-05, 3.9881e-05], device='cuda:1') +2023-03-20 18:13:41,360 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 18:13:42,679 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 +2023-03-20 18:13:45,006 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3660, 3.6609, 3.7173, 3.5238, 3.2610, 3.6256, 3.8710, 3.9952], + device='cuda:1'), covar=tensor([0.0518, 0.0294, 0.0370, 0.0389, 0.0534, 0.0462, 0.0413, 0.0245], + device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0071, 0.0066, 0.0088, 0.0072, 0.0058, 0.0060, 0.0059], + device='cuda:1'), out_proj_covar=tensor([1.1000e-04, 1.0463e-04, 9.7679e-05, 1.4639e-04, 1.1304e-04, 8.8351e-05, + 8.9608e-05, 8.8341e-05], device='cuda:1') +2023-03-20 18:13:48,032 INFO [train.py:901] (1/2) Epoch 3, batch 800, loss[loss=0.3698, simple_loss=0.3911, pruned_loss=0.1743, over 6783.00 frames. ], tot_loss[loss=0.2781, simple_loss=0.3203, pruned_loss=0.1179, over 1416969.79 frames. ], batch size: 107, lr: 3.81e-02, grad_scale: 16.0 +2023-03-20 18:13:48,046 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 18:13:49,620 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 18:13:53,255 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=6459.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:13:53,521 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-20 18:13:54,868 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6462.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:13:57,308 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6467.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:13:59,743 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 18:14:10,181 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6491.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:14:13,225 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9245, 3.7731, 3.4163, 3.8179, 3.2091, 2.6111, 3.9784, 3.5648], + device='cuda:1'), covar=tensor([0.0128, 0.0049, 0.0118, 0.0088, 0.0285, 0.1223, 0.0150, 0.0675], + device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0063, 0.0097, 0.0080, 0.0112, 0.0207, 0.0080, 0.0170], + device='cuda:1'), out_proj_covar=tensor([5.0388e-05, 4.2333e-05, 6.2196e-05, 5.3794e-05, 7.5980e-05, 1.4506e-04, + 5.6068e-05, 1.2524e-04], device='cuda:1') +2023-03-20 18:14:14,093 INFO [train.py:901] (1/2) Epoch 3, batch 850, loss[loss=0.2841, simple_loss=0.3281, pruned_loss=0.12, over 7340.00 frames. ], tot_loss[loss=0.2779, simple_loss=0.32, pruned_loss=0.1179, over 1422148.64 frames. ], batch size: 54, lr: 3.80e-02, grad_scale: 16.0 +2023-03-20 18:14:16,100 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.432e+02 3.714e+02 4.604e+02 5.774e+02 1.158e+03, threshold=9.207e+02, percent-clipped=4.0 +2023-03-20 18:14:18,160 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 18:14:18,678 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 18:14:21,904 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6514.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:14:23,837 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 18:14:26,500 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6523.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 18:14:27,325 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 18:14:27,928 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6526.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:14:29,023 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6528.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:14:40,320 INFO [train.py:901] (1/2) Epoch 3, batch 900, loss[loss=0.2749, simple_loss=0.3271, pruned_loss=0.1113, over 7226.00 frames. ], tot_loss[loss=0.2765, simple_loss=0.3189, pruned_loss=0.117, over 1426983.57 frames. ], batch size: 93, lr: 3.79e-02, grad_scale: 16.0 +2023-03-20 18:14:42,617 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 +2023-03-20 18:14:45,519 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6559.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:14:47,087 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=6562.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:14:49,673 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6567.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:15:06,118 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 18:15:06,584 INFO [train.py:901] (1/2) Epoch 3, batch 950, loss[loss=0.2642, simple_loss=0.302, pruned_loss=0.1132, over 7137.00 frames. ], tot_loss[loss=0.2761, simple_loss=0.3188, pruned_loss=0.1167, over 1431443.86 frames. ], batch size: 41, lr: 3.78e-02, grad_scale: 16.0 +2023-03-20 18:15:08,979 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 3.725e+02 4.764e+02 5.478e+02 1.056e+03, threshold=9.528e+02, percent-clipped=3.0 +2023-03-20 18:15:19,396 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6622.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:15:23,042 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6629.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:15:30,021 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 18:15:31,645 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6646.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:15:33,027 INFO [train.py:901] (1/2) Epoch 3, batch 1000, loss[loss=0.2519, simple_loss=0.2989, pruned_loss=0.1025, over 7327.00 frames. ], tot_loss[loss=0.2758, simple_loss=0.3188, pruned_loss=0.1164, over 1434446.24 frames. ], batch size: 75, lr: 3.78e-02, grad_scale: 16.0 +2023-03-20 18:15:49,704 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 18:15:54,405 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6690.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:15:56,360 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=6694.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:15:58,869 INFO [train.py:901] (1/2) Epoch 3, batch 1050, loss[loss=0.2998, simple_loss=0.3369, pruned_loss=0.1314, over 7344.00 frames. ], tot_loss[loss=0.2777, simple_loss=0.32, pruned_loss=0.1177, over 1434414.69 frames. ], batch size: 75, lr: 3.77e-02, grad_scale: 16.0 +2023-03-20 18:15:59,599 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6699.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:16:01,468 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.188e+02 3.614e+02 4.562e+02 6.271e+02 1.467e+03, threshold=9.125e+02, percent-clipped=4.0 +2023-03-20 18:16:07,359 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 +2023-03-20 18:16:11,077 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 18:16:15,172 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 18:16:15,942 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.02 vs. limit=2.0 +2023-03-20 18:16:24,532 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=6747.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:16:25,489 INFO [train.py:901] (1/2) Epoch 3, batch 1100, loss[loss=0.1945, simple_loss=0.2431, pruned_loss=0.07299, over 6486.00 frames. ], tot_loss[loss=0.2752, simple_loss=0.3186, pruned_loss=0.1159, over 1436370.07 frames. ], batch size: 28, lr: 3.76e-02, grad_scale: 16.0 +2023-03-20 18:16:43,764 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 18:16:44,244 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 18:16:47,423 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6791.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:16:51,357 INFO [train.py:901] (1/2) Epoch 3, batch 1150, loss[loss=0.2662, simple_loss=0.3137, pruned_loss=0.1093, over 7264.00 frames. ], tot_loss[loss=0.2741, simple_loss=0.3177, pruned_loss=0.1153, over 1437124.51 frames. ], batch size: 52, lr: 3.75e-02, grad_scale: 16.0 +2023-03-20 18:16:53,753 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.995e+02 3.595e+02 4.369e+02 5.397e+02 1.197e+03, threshold=8.738e+02, percent-clipped=1.0 +2023-03-20 18:16:57,414 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 18:16:57,902 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 18:17:01,536 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6818.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 18:17:03,973 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6823.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:17:05,511 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6826.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:17:12,658 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=6839.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:17:17,593 INFO [train.py:901] (1/2) Epoch 3, batch 1200, loss[loss=0.3262, simple_loss=0.3524, pruned_loss=0.15, over 7266.00 frames. ], tot_loss[loss=0.2746, simple_loss=0.3175, pruned_loss=0.1158, over 1437375.33 frames. ], batch size: 77, lr: 3.74e-02, grad_scale: 16.0 +2023-03-20 18:17:22,723 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6859.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:17:27,410 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6867.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:17:30,878 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=6874.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:17:31,326 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 18:17:33,004 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6878.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 18:17:43,501 INFO [train.py:901] (1/2) Epoch 3, batch 1250, loss[loss=0.2954, simple_loss=0.3327, pruned_loss=0.1291, over 7273.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3175, pruned_loss=0.1159, over 1436098.13 frames. ], batch size: 70, lr: 3.73e-02, grad_scale: 16.0 +2023-03-20 18:17:45,505 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.075e+02 4.156e+02 4.876e+02 5.701e+02 1.197e+03, threshold=9.752e+02, percent-clipped=5.0 +2023-03-20 18:17:47,641 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=6907.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:17:48,735 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6909.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:17:51,817 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=6915.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:17:54,814 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 18:17:56,069 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6922.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:17:59,546 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 18:18:01,080 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 18:18:04,770 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6939.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 18:18:09,649 INFO [train.py:901] (1/2) Epoch 3, batch 1300, loss[loss=0.3329, simple_loss=0.3737, pruned_loss=0.1461, over 7116.00 frames. ], tot_loss[loss=0.2733, simple_loss=0.3165, pruned_loss=0.1151, over 1435451.86 frames. ], batch size: 98, lr: 3.72e-02, grad_scale: 16.0 +2023-03-20 18:18:20,968 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:18:21,037 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:18:24,463 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 18:18:26,538 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 18:18:28,629 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6985.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:18:29,601 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 18:18:29,736 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3558, 3.4884, 1.8495, 3.3437, 3.1974, 2.9068, 3.1936, 2.3488], + device='cuda:1'), covar=tensor([0.1125, 0.0164, 0.0278, 0.0103, 0.0286, 0.0717, 0.0170, 0.0369], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0025, 0.0027, 0.0025, 0.0025, 0.0025, 0.0025, 0.0024], + device='cuda:1'), out_proj_covar=tensor([3.9302e-05, 3.2901e-05, 3.8692e-05, 3.4081e-05, 3.6022e-05, 3.7500e-05, + 3.4248e-05, 3.9008e-05], device='cuda:1') +2023-03-20 18:18:35,688 INFO [train.py:901] (1/2) Epoch 3, batch 1350, loss[loss=0.2313, simple_loss=0.2778, pruned_loss=0.09236, over 7019.00 frames. ], tot_loss[loss=0.2715, simple_loss=0.3152, pruned_loss=0.1139, over 1436693.00 frames. ], batch size: 35, lr: 3.71e-02, grad_scale: 16.0 +2023-03-20 18:18:37,961 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.522e+02 3.667e+02 4.621e+02 5.823e+02 1.455e+03, threshold=9.242e+02, percent-clipped=1.0 +2023-03-20 18:18:40,699 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 18:19:02,193 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7048.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 18:19:02,501 INFO [train.py:901] (1/2) Epoch 3, batch 1400, loss[loss=0.312, simple_loss=0.3398, pruned_loss=0.1421, over 7255.00 frames. ], tot_loss[loss=0.271, simple_loss=0.3149, pruned_loss=0.1135, over 1438748.99 frames. ], batch size: 47, lr: 3.70e-02, grad_scale: 16.0 +2023-03-20 18:19:09,682 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8563, 3.7698, 3.6471, 3.1271, 3.7311, 3.4534, 1.6982, 3.9096], + device='cuda:1'), covar=tensor([0.0037, 0.0087, 0.0067, 0.0219, 0.0051, 0.0253, 0.1441, 0.0075], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0031, 0.0040, 0.0034, 0.0034, 0.0045, 0.0071, 0.0036], + device='cuda:1'), out_proj_covar=tensor([3.5617e-05, 4.3767e-05, 4.9141e-05, 4.6567e-05, 3.7217e-05, 5.9706e-05, + 1.0010e-04, 4.5228e-05], device='cuda:1') +2023-03-20 18:19:13,006 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.00 vs. limit=2.0 +2023-03-20 18:19:13,208 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 18:19:26,764 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7262, 2.6619, 2.1178, 2.7597, 2.7878, 2.7331, 2.2494, 1.7348], + device='cuda:1'), covar=tensor([0.0064, 0.0247, 0.0819, 0.0443, 0.0152, 0.0160, 0.0843, 0.1028], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0061, 0.0120, 0.0065, 0.0060, 0.0054, 0.0130, 0.0121], + device='cuda:1'), out_proj_covar=tensor([4.7003e-05, 6.0242e-05, 1.0997e-04, 6.0474e-05, 5.6126e-05, 5.1470e-05, + 1.2230e-04, 1.0912e-04], device='cuda:1') +2023-03-20 18:19:28,590 INFO [train.py:901] (1/2) Epoch 3, batch 1450, loss[loss=0.2621, simple_loss=0.302, pruned_loss=0.1111, over 7337.00 frames. ], tot_loss[loss=0.2702, simple_loss=0.3147, pruned_loss=0.1128, over 1440496.38 frames. ], batch size: 51, lr: 3.70e-02, grad_scale: 16.0 +2023-03-20 18:19:30,650 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.276e+02 3.678e+02 4.650e+02 5.818e+02 1.874e+03, threshold=9.299e+02, percent-clipped=3.0 +2023-03-20 18:19:33,907 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7109.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 18:19:38,416 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7118.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 18:19:38,806 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 18:19:41,472 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7123.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:19:54,153 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 18:19:54,633 INFO [train.py:901] (1/2) Epoch 3, batch 1500, loss[loss=0.2733, simple_loss=0.3228, pruned_loss=0.1119, over 7127.00 frames. ], tot_loss[loss=0.27, simple_loss=0.3147, pruned_loss=0.1127, over 1441971.46 frames. ], batch size: 98, lr: 3.69e-02, grad_scale: 16.0 +2023-03-20 18:19:58,717 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5518, 4.6553, 4.8234, 4.8575, 4.6242, 4.2093, 4.9451, 4.8021], + device='cuda:1'), covar=tensor([0.0359, 0.0258, 0.0246, 0.0451, 0.0353, 0.0367, 0.0283, 0.0324], + device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0059, 0.0074, 0.0066, 0.0056, 0.0067, 0.0061, 0.0058], + device='cuda:1'), out_proj_covar=tensor([1.0237e-04, 7.8678e-05, 1.0501e-04, 9.8142e-05, 8.0109e-05, 9.5899e-05, + 8.6230e-05, 8.1858e-05], device='cuda:1') +2023-03-20 18:20:01,842 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7163.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:20:03,299 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=7166.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:20:06,373 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=7171.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:20:14,489 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1861, 1.9071, 2.1516, 2.1756, 1.9542, 1.7601, 2.3928, 1.5936], + device='cuda:1'), covar=tensor([0.0440, 0.1119, 0.0614, 0.0324, 0.0789, 0.0555, 0.0550, 0.1328], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0037, 0.0037, 0.0038, 0.0034, 0.0039, 0.0040, 0.0036], + device='cuda:1'), out_proj_covar=tensor([6.1567e-05, 5.1715e-05, 5.1901e-05, 5.1415e-05, 4.8472e-05, 5.5192e-05, + 5.7447e-05, 5.1486e-05], device='cuda:1') +2023-03-20 18:20:17,889 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 18:20:20,399 INFO [train.py:901] (1/2) Epoch 3, batch 1550, loss[loss=0.2745, simple_loss=0.3268, pruned_loss=0.1111, over 7242.00 frames. ], tot_loss[loss=0.2689, simple_loss=0.3137, pruned_loss=0.112, over 1440693.08 frames. ], batch size: 93, lr: 3.68e-02, grad_scale: 16.0 +2023-03-20 18:20:22,739 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.058e+02 3.898e+02 4.822e+02 6.502e+02 1.544e+03, threshold=9.644e+02, percent-clipped=7.0 +2023-03-20 18:20:34,353 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7224.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 18:20:39,243 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7234.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 18:20:45,732 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0619, 3.9830, 3.9995, 4.4311, 4.5119, 4.4443, 3.9708, 3.7785], + device='cuda:1'), covar=tensor([0.0753, 0.1497, 0.1875, 0.1126, 0.0488, 0.0916, 0.0558, 0.0778], + device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0145, 0.0147, 0.0120, 0.0100, 0.0133, 0.0083, 0.0103], + device='cuda:1'), out_proj_covar=tensor([8.9508e-05, 1.6272e-04, 1.6981e-04, 1.4628e-04, 1.0620e-04, 1.5701e-04, + 8.8484e-05, 1.0692e-04], device='cuda:1') +2023-03-20 18:20:46,674 INFO [train.py:901] (1/2) Epoch 3, batch 1600, loss[loss=0.2175, simple_loss=0.2612, pruned_loss=0.08689, over 6991.00 frames. ], tot_loss[loss=0.2687, simple_loss=0.3133, pruned_loss=0.1121, over 1438976.14 frames. ], batch size: 35, lr: 3.67e-02, grad_scale: 16.0 +2023-03-20 18:20:50,242 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 18:20:50,772 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 18:20:51,631 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.49 vs. limit=5.0 +2023-03-20 18:20:53,810 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 18:20:55,492 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7265.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:21:00,631 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7275.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:21:04,587 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 18:21:05,707 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7285.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:21:08,680 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 18:21:13,303 INFO [train.py:901] (1/2) Epoch 3, batch 1650, loss[loss=0.2907, simple_loss=0.3326, pruned_loss=0.1244, over 7141.00 frames. ], tot_loss[loss=0.2693, simple_loss=0.3133, pruned_loss=0.1127, over 1436881.43 frames. ], batch size: 98, lr: 3.66e-02, grad_scale: 16.0 +2023-03-20 18:21:15,324 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.873e+02 4.152e+02 5.169e+02 6.417e+02 1.445e+03, threshold=1.034e+03, percent-clipped=2.0 +2023-03-20 18:21:17,390 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 18:21:29,641 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4505, 4.4887, 4.4973, 4.9638, 5.0472, 5.1126, 4.4836, 4.4352], + device='cuda:1'), covar=tensor([0.0837, 0.1773, 0.1928, 0.1360, 0.0544, 0.1006, 0.0617, 0.0825], + device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0151, 0.0155, 0.0127, 0.0104, 0.0139, 0.0085, 0.0106], + device='cuda:1'), out_proj_covar=tensor([9.4295e-05, 1.6939e-04, 1.7688e-04, 1.5363e-04, 1.1041e-04, 1.6430e-04, + 9.0891e-05, 1.1020e-04], device='cuda:1') +2023-03-20 18:21:30,688 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=7333.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:21:32,328 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7336.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:21:35,773 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 18:21:37,578 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 +2023-03-20 18:21:38,773 INFO [train.py:901] (1/2) Epoch 3, batch 1700, loss[loss=0.2765, simple_loss=0.3211, pruned_loss=0.1159, over 7259.00 frames. ], tot_loss[loss=0.2696, simple_loss=0.3136, pruned_loss=0.1129, over 1438586.38 frames. ], batch size: 64, lr: 3.65e-02, grad_scale: 16.0 +2023-03-20 18:21:39,793 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 18:21:51,228 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 18:22:04,197 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-20 18:22:05,302 INFO [train.py:901] (1/2) Epoch 3, batch 1750, loss[loss=0.2852, simple_loss=0.3269, pruned_loss=0.1218, over 7251.00 frames. ], tot_loss[loss=0.2698, simple_loss=0.3137, pruned_loss=0.113, over 1441107.43 frames. ], batch size: 89, lr: 3.64e-02, grad_scale: 16.0 +2023-03-20 18:22:07,665 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.208e+02 3.884e+02 4.725e+02 5.852e+02 1.126e+03, threshold=9.450e+02, percent-clipped=2.0 +2023-03-20 18:22:08,245 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7404.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 18:22:08,831 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6376, 3.3953, 3.1975, 3.4587, 2.7959, 2.3050, 3.5367, 3.1600], + device='cuda:1'), covar=tensor([0.0133, 0.0056, 0.0180, 0.0064, 0.0298, 0.1100, 0.0170, 0.0708], + device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0068, 0.0107, 0.0086, 0.0124, 0.0207, 0.0082, 0.0184], + device='cuda:1'), out_proj_covar=tensor([5.6994e-05, 4.9546e-05, 7.2831e-05, 5.8299e-05, 8.9110e-05, 1.5079e-04, + 6.1403e-05, 1.3867e-04], device='cuda:1') +2023-03-20 18:22:09,782 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7407.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:22:16,240 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 18:22:17,251 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 18:22:22,485 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.9354, 1.2106, 2.3653, 1.3352, 0.7732, 1.2013, 2.1264, 1.2503], + device='cuda:1'), covar=tensor([0.0958, 0.0874, 0.0189, 0.0433, 0.1088, 0.1113, 0.0412, 0.1196], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0037, 0.0031, 0.0031, 0.0028, 0.0030, 0.0033, 0.0033], + device='cuda:1'), out_proj_covar=tensor([3.8820e-05, 4.9624e-05, 4.1769e-05, 3.8399e-05, 3.8809e-05, 4.0289e-05, + 3.5883e-05, 4.0945e-05], device='cuda:1') +2023-03-20 18:22:31,642 INFO [train.py:901] (1/2) Epoch 3, batch 1800, loss[loss=0.2592, simple_loss=0.3027, pruned_loss=0.1079, over 7137.00 frames. ], tot_loss[loss=0.2686, simple_loss=0.3131, pruned_loss=0.1121, over 1440143.35 frames. ], batch size: 41, lr: 3.64e-02, grad_scale: 32.0 +2023-03-20 18:22:35,313 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0570, 1.0835, 0.9510, 0.8700, 1.8199, 0.6041, 0.7708, 0.8791], + device='cuda:1'), covar=tensor([0.0299, 0.0190, 0.0522, 0.0617, 0.0213, 0.0522, 0.0488, 0.0376], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0024, 0.0022, 0.0024, 0.0022, 0.0025, 0.0024, 0.0024], + device='cuda:1'), out_proj_covar=tensor([2.5805e-05, 2.3315e-05, 2.5859e-05, 2.8706e-05, 2.2387e-05, 2.8703e-05, + 3.0870e-05, 2.8243e-05], device='cuda:1') +2023-03-20 18:22:40,227 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 18:22:41,374 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7468.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:22:52,885 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 18:22:56,755 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-20 18:22:57,425 INFO [train.py:901] (1/2) Epoch 3, batch 1850, loss[loss=0.2526, simple_loss=0.3074, pruned_loss=0.0989, over 7359.00 frames. ], tot_loss[loss=0.2695, simple_loss=0.3138, pruned_loss=0.1125, over 1442256.82 frames. ], batch size: 63, lr: 3.63e-02, grad_scale: 32.0 +2023-03-20 18:22:59,427 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.848e+02 3.617e+02 4.496e+02 5.652e+02 1.259e+03, threshold=8.993e+02, percent-clipped=2.0 +2023-03-20 18:23:02,460 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 18:23:07,615 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7519.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 18:23:14,526 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-03-20 18:23:15,725 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7534.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 18:23:17,860 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0401, 2.6949, 2.0539, 2.9949, 1.7528, 3.1962, 2.7327, 3.0979], + device='cuda:1'), covar=tensor([0.0042, 0.0531, 0.2531, 0.0047, 0.4894, 0.0070, 0.0585, 0.0077], + device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0133, 0.0244, 0.0087, 0.0224, 0.0089, 0.0157, 0.0090], + device='cuda:1'), out_proj_covar=tensor([6.7601e-05, 1.1083e-04, 1.8260e-04, 7.1194e-05, 1.7609e-04, 7.1629e-05, + 1.2731e-04, 7.2947e-05], device='cuda:1') +2023-03-20 18:23:19,689 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 18:23:23,222 INFO [train.py:901] (1/2) Epoch 3, batch 1900, loss[loss=0.2707, simple_loss=0.3185, pruned_loss=0.1114, over 7354.00 frames. ], tot_loss[loss=0.2697, simple_loss=0.3135, pruned_loss=0.1129, over 1441799.53 frames. ], batch size: 63, lr: 3.62e-02, grad_scale: 32.0 +2023-03-20 18:23:25,533 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.95 vs. limit=5.0 +2023-03-20 18:23:31,908 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7565.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:23:40,262 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=7582.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 18:23:44,720 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 18:23:48,790 INFO [train.py:901] (1/2) Epoch 3, batch 1950, loss[loss=0.2843, simple_loss=0.3299, pruned_loss=0.1193, over 7257.00 frames. ], tot_loss[loss=0.269, simple_loss=0.3132, pruned_loss=0.1125, over 1443611.91 frames. ], batch size: 52, lr: 3.61e-02, grad_scale: 32.0 +2023-03-20 18:23:51,158 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.297e+02 4.106e+02 4.807e+02 6.855e+02 1.279e+03, threshold=9.614e+02, percent-clipped=5.0 +2023-03-20 18:23:56,325 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 18:23:56,374 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=7613.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:24:02,039 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 18:24:02,551 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 18:24:02,643 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7624.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:24:06,096 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7631.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:24:15,186 INFO [train.py:901] (1/2) Epoch 3, batch 2000, loss[loss=0.2806, simple_loss=0.32, pruned_loss=0.1206, over 7329.00 frames. ], tot_loss[loss=0.2681, simple_loss=0.3123, pruned_loss=0.1119, over 1443427.03 frames. ], batch size: 59, lr: 3.60e-02, grad_scale: 32.0 +2023-03-20 18:24:20,428 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 18:24:25,171 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8355, 2.7530, 2.1150, 2.6542, 1.7055, 2.8621, 2.4254, 2.7491], + device='cuda:1'), covar=tensor([0.0136, 0.0476, 0.2972, 0.0068, 0.6887, 0.0062, 0.0830, 0.0098], + device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0138, 0.0253, 0.0088, 0.0233, 0.0091, 0.0161, 0.0093], + device='cuda:1'), out_proj_covar=tensor([7.1733e-05, 1.1485e-04, 1.8996e-04, 7.1694e-05, 1.8339e-04, 7.2886e-05, + 1.3084e-04, 7.5202e-05], device='cuda:1') +2023-03-20 18:24:31,543 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 18:24:34,162 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7685.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:24:39,205 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 18:24:41,213 INFO [train.py:901] (1/2) Epoch 3, batch 2050, loss[loss=0.2478, simple_loss=0.3015, pruned_loss=0.09708, over 7329.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.3117, pruned_loss=0.1113, over 1443959.73 frames. ], batch size: 61, lr: 3.59e-02, grad_scale: 32.0 +2023-03-20 18:24:43,220 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.099e+02 3.470e+02 4.322e+02 5.464e+02 1.086e+03, threshold=8.645e+02, percent-clipped=2.0 +2023-03-20 18:24:43,790 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7704.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 18:24:48,297 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-03-20 18:24:53,219 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9819, 1.9673, 2.3393, 2.7777, 2.5689, 2.7346, 2.4411, 2.9738], + device='cuda:1'), covar=tensor([0.1601, 0.0908, 0.1988, 0.0231, 0.0058, 0.0053, 0.0052, 0.0077], + device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0125, 0.0169, 0.0081, 0.0066, 0.0073, 0.0068, 0.0070], + device='cuda:1'), out_proj_covar=tensor([1.4265e-04, 1.0658e-04, 1.4096e-04, 7.5538e-05, 5.5182e-05, 6.0317e-05, + 5.9807e-05, 5.6680e-05], device='cuda:1') +2023-03-20 18:25:04,857 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7743.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:25:07,721 INFO [train.py:901] (1/2) Epoch 3, batch 2100, loss[loss=0.3046, simple_loss=0.3433, pruned_loss=0.133, over 7203.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3115, pruned_loss=0.111, over 1444917.98 frames. ], batch size: 93, lr: 3.59e-02, grad_scale: 16.0 +2023-03-20 18:25:09,333 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=7752.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 18:25:14,210 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 18:25:14,817 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7763.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:25:17,256 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 18:25:22,530 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6384, 2.0637, 2.1545, 2.5559, 2.4747, 2.4320, 2.1402, 2.4060], + device='cuda:1'), covar=tensor([0.0352, 0.0122, 0.0510, 0.0595, 0.0456, 0.0609, 0.0578, 0.0443], + device='cuda:1'), in_proj_covar=tensor([0.0026, 0.0023, 0.0026, 0.0029, 0.0029, 0.0028, 0.0030, 0.0026], + device='cuda:1'), out_proj_covar=tensor([3.9866e-05, 3.4131e-05, 4.2152e-05, 4.5941e-05, 4.4372e-05, 4.3713e-05, + 4.5898e-05, 4.2960e-05], device='cuda:1') +2023-03-20 18:25:26,565 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3143, 4.4375, 4.3908, 4.7907, 4.9008, 4.8416, 4.4250, 4.2530], + device='cuda:1'), covar=tensor([0.0824, 0.1322, 0.1770, 0.1212, 0.0538, 0.1110, 0.0534, 0.0823], + device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0146, 0.0148, 0.0123, 0.0101, 0.0141, 0.0085, 0.0103], + device='cuda:1'), out_proj_covar=tensor([9.5608e-05, 1.6548e-04, 1.6817e-04, 1.4851e-04, 1.0763e-04, 1.6713e-04, + 9.4408e-05, 1.0808e-04], device='cuda:1') +2023-03-20 18:25:33,569 INFO [train.py:901] (1/2) Epoch 3, batch 2150, loss[loss=0.304, simple_loss=0.3295, pruned_loss=0.1392, over 7269.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3104, pruned_loss=0.1103, over 1443842.20 frames. ], batch size: 89, lr: 3.58e-02, grad_scale: 16.0 +2023-03-20 18:25:35,565 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7802.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:25:36,408 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.942e+02 3.904e+02 5.108e+02 6.529e+02 1.416e+03, threshold=1.022e+03, percent-clipped=8.0 +2023-03-20 18:25:36,683 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7804.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:25:44,174 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7819.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 18:25:59,739 INFO [train.py:901] (1/2) Epoch 3, batch 2200, loss[loss=0.2749, simple_loss=0.3159, pruned_loss=0.1169, over 7317.00 frames. ], tot_loss[loss=0.267, simple_loss=0.312, pruned_loss=0.111, over 1443748.09 frames. ], batch size: 59, lr: 3.57e-02, grad_scale: 16.0 +2023-03-20 18:26:02,287 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 18:26:07,004 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7863.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:26:08,897 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=7867.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:26:09,972 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5605, 3.8654, 3.5480, 3.2382, 3.5342, 3.0228, 1.1386, 3.6898], + device='cuda:1'), covar=tensor([0.0044, 0.0093, 0.0085, 0.0140, 0.0047, 0.0342, 0.1823, 0.0123], + device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0033, 0.0043, 0.0035, 0.0038, 0.0050, 0.0075, 0.0040], + device='cuda:1'), out_proj_covar=tensor([3.9662e-05, 4.8826e-05, 5.6648e-05, 4.8993e-05, 4.5625e-05, 6.8452e-05, + 1.0773e-04, 5.2092e-05], device='cuda:1') +2023-03-20 18:26:25,703 INFO [train.py:901] (1/2) Epoch 3, batch 2250, loss[loss=0.3224, simple_loss=0.3556, pruned_loss=0.1446, over 7213.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3115, pruned_loss=0.1109, over 1443342.62 frames. ], batch size: 93, lr: 3.56e-02, grad_scale: 16.0 +2023-03-20 18:26:26,842 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3394, 3.5827, 3.2428, 3.5527, 3.3997, 3.4956, 3.5812, 3.1023], + device='cuda:1'), covar=tensor([0.0092, 0.0115, 0.0153, 0.0123, 0.0116, 0.0115, 0.0085, 0.0154], + device='cuda:1'), in_proj_covar=tensor([0.0022, 0.0023, 0.0021, 0.0021, 0.0020, 0.0023, 0.0025, 0.0023], + device='cuda:1'), out_proj_covar=tensor([4.7606e-05, 5.6397e-05, 5.0280e-05, 4.7723e-05, 4.8323e-05, 5.0536e-05, + 5.8567e-05, 5.2391e-05], device='cuda:1') +2023-03-20 18:26:28,198 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.204e+02 3.580e+02 4.617e+02 6.250e+02 1.322e+03, threshold=9.234e+02, percent-clipped=2.0 +2023-03-20 18:26:29,662 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 +2023-03-20 18:26:36,924 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 18:26:37,408 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 18:26:42,580 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7931.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:26:50,012 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 18:26:51,457 INFO [train.py:901] (1/2) Epoch 3, batch 2300, loss[loss=0.2766, simple_loss=0.3238, pruned_loss=0.1146, over 7273.00 frames. ], tot_loss[loss=0.2671, simple_loss=0.3122, pruned_loss=0.1111, over 1445917.45 frames. ], batch size: 57, lr: 3.55e-02, grad_scale: 16.0 +2023-03-20 18:26:52,304 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 +2023-03-20 18:26:52,951 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.29 vs. limit=2.0 +2023-03-20 18:26:55,908 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 +2023-03-20 18:26:59,796 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.20 vs. limit=2.0 +2023-03-20 18:27:07,547 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=7979.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:27:08,057 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7980.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:27:18,150 INFO [train.py:901] (1/2) Epoch 3, batch 2350, loss[loss=0.2635, simple_loss=0.3086, pruned_loss=0.1092, over 7343.00 frames. ], tot_loss[loss=0.2665, simple_loss=0.3121, pruned_loss=0.1105, over 1446796.53 frames. ], batch size: 54, lr: 3.55e-02, grad_scale: 16.0 +2023-03-20 18:27:24,519 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.608e+02 3.747e+02 4.691e+02 6.358e+02 1.596e+03, threshold=9.383e+02, percent-clipped=3.0 +2023-03-20 18:27:29,703 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1946, 2.3501, 1.0113, 2.2150, 1.1806, 1.6570, 1.8111, 1.2320], + device='cuda:1'), covar=tensor([0.0473, 0.0206, 0.0479, 0.0177, 0.0624, 0.0543, 0.0328, 0.0864], + device='cuda:1'), in_proj_covar=tensor([0.0022, 0.0024, 0.0024, 0.0024, 0.0023, 0.0023, 0.0024, 0.0023], + device='cuda:1'), out_proj_covar=tensor([3.7651e-05, 3.6172e-05, 3.7230e-05, 3.6014e-05, 3.7944e-05, 3.8521e-05, + 3.7711e-05, 4.0282e-05], device='cuda:1') +2023-03-20 18:27:33,289 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0282, 0.8777, 0.7114, 0.6761, 1.4361, 1.0504, 1.0879, 0.6601], + device='cuda:1'), covar=tensor([0.0349, 0.0229, 0.0682, 0.0568, 0.0159, 0.0240, 0.0184, 0.0538], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0023, 0.0020, 0.0019, 0.0020, 0.0021, 0.0022, 0.0021], + device='cuda:1'), out_proj_covar=tensor([2.4060e-05, 2.2658e-05, 2.4063e-05, 2.3354e-05, 2.0390e-05, 2.4906e-05, + 2.8292e-05, 2.5232e-05], device='cuda:1') +2023-03-20 18:27:38,049 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 18:27:39,656 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8347, 3.4927, 3.4904, 3.6081, 2.8705, 2.4014, 3.7167, 3.1149], + device='cuda:1'), covar=tensor([0.0136, 0.0057, 0.0162, 0.0041, 0.0255, 0.0622, 0.0107, 0.0611], + device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0076, 0.0123, 0.0090, 0.0146, 0.0222, 0.0094, 0.0203], + device='cuda:1'), out_proj_covar=tensor([6.8343e-05, 5.7864e-05, 8.7758e-05, 6.3587e-05, 1.0813e-04, 1.6435e-04, + 7.2813e-05, 1.5508e-04], device='cuda:1') +2023-03-20 18:27:40,079 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6943, 4.7945, 5.0412, 5.0495, 4.8898, 4.5037, 5.1912, 4.9303], + device='cuda:1'), covar=tensor([0.0349, 0.0219, 0.0327, 0.0468, 0.0301, 0.0265, 0.0247, 0.0365], + device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0062, 0.0079, 0.0066, 0.0057, 0.0070, 0.0063, 0.0059], + device='cuda:1'), out_proj_covar=tensor([1.1319e-04, 8.4964e-05, 1.1735e-04, 1.0098e-04, 8.3825e-05, 9.9896e-05, + 9.0938e-05, 8.5523e-05], device='cuda:1') +2023-03-20 18:27:46,319 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 18:27:47,815 INFO [train.py:901] (1/2) Epoch 3, batch 2400, loss[loss=0.2066, simple_loss=0.245, pruned_loss=0.08408, over 6516.00 frames. ], tot_loss[loss=0.2681, simple_loss=0.3133, pruned_loss=0.1115, over 1444385.86 frames. ], batch size: 28, lr: 3.54e-02, grad_scale: 16.0 +2023-03-20 18:27:54,932 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8063.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:27:56,930 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 18:27:59,560 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 18:28:04,469 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2974, 2.3176, 2.2416, 1.9029, 1.8541, 1.7161, 2.4599, 1.8386], + device='cuda:1'), covar=tensor([0.0565, 0.0252, 0.0577, 0.0571, 0.0571, 0.0850, 0.0283, 0.1436], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0034, 0.0037, 0.0040, 0.0032, 0.0036, 0.0039, 0.0039], + device='cuda:1'), out_proj_covar=tensor([6.4926e-05, 5.1315e-05, 5.8235e-05, 5.8115e-05, 5.0905e-05, 5.5752e-05, + 6.2356e-05, 5.9798e-05], device='cuda:1') +2023-03-20 18:28:09,716 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.10 vs. limit=2.0 +2023-03-20 18:28:13,980 INFO [train.py:901] (1/2) Epoch 3, batch 2450, loss[loss=0.2679, simple_loss=0.3176, pruned_loss=0.1091, over 7354.00 frames. ], tot_loss[loss=0.2691, simple_loss=0.3141, pruned_loss=0.1121, over 1443469.28 frames. ], batch size: 54, lr: 3.53e-02, grad_scale: 16.0 +2023-03-20 18:28:14,073 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8099.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:28:16,490 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.331e+02 4.135e+02 4.933e+02 6.435e+02 1.136e+03, threshold=9.866e+02, percent-clipped=3.0 +2023-03-20 18:28:20,176 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=8111.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:28:20,756 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8160, 2.2602, 2.0262, 1.6633, 2.1974, 2.2184, 1.9230, 2.4092], + device='cuda:1'), covar=tensor([0.0673, 0.0138, 0.0695, 0.1554, 0.0534, 0.0850, 0.0678, 0.0525], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0022, 0.0026, 0.0029, 0.0027, 0.0027, 0.0030, 0.0025], + device='cuda:1'), out_proj_covar=tensor([4.0463e-05, 3.3002e-05, 4.2943e-05, 4.7129e-05, 4.2662e-05, 4.2809e-05, + 4.8517e-05, 4.2679e-05], device='cuda:1') +2023-03-20 18:28:26,226 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 18:28:38,213 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.6853, 0.7458, 0.6211, 0.6576, 1.5311, 1.1555, 0.7950, 0.7437], + device='cuda:1'), covar=tensor([0.0392, 0.0277, 0.0559, 0.0823, 0.0210, 0.0362, 0.0980, 0.0697], + device='cuda:1'), in_proj_covar=tensor([0.0022, 0.0023, 0.0020, 0.0021, 0.0021, 0.0022, 0.0025, 0.0022], + device='cuda:1'), out_proj_covar=tensor([2.4360e-05, 2.2699e-05, 2.3907e-05, 2.4649e-05, 2.1005e-05, 2.5636e-05, + 3.1461e-05, 2.7107e-05], device='cuda:1') +2023-03-20 18:28:40,170 INFO [train.py:901] (1/2) Epoch 3, batch 2500, loss[loss=0.2903, simple_loss=0.3272, pruned_loss=0.1266, over 7329.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.312, pruned_loss=0.1107, over 1443949.09 frames. ], batch size: 61, lr: 3.52e-02, grad_scale: 16.0 +2023-03-20 18:28:44,802 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8158.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:28:53,945 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 18:28:59,436 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-20 18:29:06,260 INFO [train.py:901] (1/2) Epoch 3, batch 2550, loss[loss=0.2465, simple_loss=0.2986, pruned_loss=0.0972, over 7297.00 frames. ], tot_loss[loss=0.2651, simple_loss=0.3107, pruned_loss=0.1098, over 1444595.69 frames. ], batch size: 66, lr: 3.51e-02, grad_scale: 16.0 +2023-03-20 18:29:09,135 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.629e+02 3.850e+02 5.119e+02 6.158e+02 1.209e+03, threshold=1.024e+03, percent-clipped=1.0 +2023-03-20 18:29:32,413 INFO [train.py:901] (1/2) Epoch 3, batch 2600, loss[loss=0.251, simple_loss=0.2969, pruned_loss=0.1026, over 7243.00 frames. ], tot_loss[loss=0.2644, simple_loss=0.3101, pruned_loss=0.1093, over 1443728.66 frames. ], batch size: 47, lr: 3.51e-02, grad_scale: 16.0 +2023-03-20 18:29:35,746 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.12 vs. limit=2.0 +2023-03-20 18:29:39,819 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-20 18:29:40,643 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8265.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:29:48,050 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8280.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:29:58,027 INFO [train.py:901] (1/2) Epoch 3, batch 2650, loss[loss=0.282, simple_loss=0.3266, pruned_loss=0.1187, over 7312.00 frames. ], tot_loss[loss=0.2627, simple_loss=0.3085, pruned_loss=0.1084, over 1443240.99 frames. ], batch size: 80, lr: 3.50e-02, grad_scale: 16.0 +2023-03-20 18:30:00,547 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.150e+02 3.672e+02 4.487e+02 5.762e+02 1.165e+03, threshold=8.975e+02, percent-clipped=3.0 +2023-03-20 18:30:11,760 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8326.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:30:12,693 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=8328.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:30:23,014 INFO [train.py:901] (1/2) Epoch 3, batch 2700, loss[loss=0.2858, simple_loss=0.329, pruned_loss=0.1213, over 7345.00 frames. ], tot_loss[loss=0.2634, simple_loss=0.309, pruned_loss=0.1089, over 1441900.28 frames. ], batch size: 73, lr: 3.49e-02, grad_scale: 16.0 +2023-03-20 18:30:48,009 INFO [train.py:901] (1/2) Epoch 3, batch 2750, loss[loss=0.2925, simple_loss=0.3329, pruned_loss=0.126, over 7362.00 frames. ], tot_loss[loss=0.263, simple_loss=0.309, pruned_loss=0.1085, over 1442355.98 frames. ], batch size: 51, lr: 3.48e-02, grad_scale: 16.0 +2023-03-20 18:30:48,102 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8399.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:30:50,824 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.974e+02 3.728e+02 5.076e+02 6.308e+02 1.197e+03, threshold=1.015e+03, percent-clipped=5.0 +2023-03-20 18:30:51,181 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.24 vs. limit=2.0 +2023-03-20 18:30:56,029 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 +2023-03-20 18:31:02,185 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2608, 3.5409, 3.4070, 3.5503, 3.6109, 3.4477, 3.7037, 3.2005], + device='cuda:1'), covar=tensor([0.0110, 0.0115, 0.0108, 0.0104, 0.0087, 0.0115, 0.0081, 0.0120], + device='cuda:1'), in_proj_covar=tensor([0.0022, 0.0024, 0.0023, 0.0021, 0.0019, 0.0023, 0.0025, 0.0024], + device='cuda:1'), out_proj_covar=tensor([5.2651e-05, 6.0537e-05, 5.7104e-05, 4.9884e-05, 4.8255e-05, 5.3695e-05, + 6.4336e-05, 5.6970e-05], device='cuda:1') +2023-03-20 18:31:12,355 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=8447.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:31:13,266 INFO [train.py:901] (1/2) Epoch 3, batch 2800, loss[loss=0.2974, simple_loss=0.3442, pruned_loss=0.1253, over 7227.00 frames. ], tot_loss[loss=0.2638, simple_loss=0.3093, pruned_loss=0.1091, over 1439789.60 frames. ], batch size: 99, lr: 3.48e-02, grad_scale: 16.0 +2023-03-20 18:31:17,823 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8458.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:31:20,263 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6461, 3.8587, 3.7144, 3.2596, 3.9316, 2.8217, 1.2946, 3.8667], + device='cuda:1'), covar=tensor([0.0052, 0.0096, 0.0112, 0.0120, 0.0037, 0.0430, 0.1460, 0.0093], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0036, 0.0045, 0.0039, 0.0039, 0.0054, 0.0076, 0.0040], + device='cuda:1'), out_proj_covar=tensor([4.6250e-05, 5.4698e-05, 5.9688e-05, 5.4361e-05, 4.5314e-05, 7.5662e-05, + 1.1183e-04, 5.1836e-05], device='cuda:1') +2023-03-20 18:31:38,791 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 18:31:38,843 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 18:31:38,868 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 18:31:38,885 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 18:31:39,059 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 18:31:39,074 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 18:31:39,085 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 18:31:39,499 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 18:31:39,573 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 18:31:39,588 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 18:31:39,604 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 18:31:39,763 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 18:31:39,769 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 18:31:39,776 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 18:31:39,914 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 18:31:39,916 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 18:31:39,974 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 18:31:48,089 INFO [train.py:901] (1/2) Epoch 4, batch 0, loss[loss=0.2814, simple_loss=0.3281, pruned_loss=0.1173, over 7259.00 frames. ], tot_loss[loss=0.2814, simple_loss=0.3281, pruned_loss=0.1173, over 7259.00 frames. ], batch size: 55, lr: 3.37e-02, grad_scale: 16.0 +2023-03-20 18:31:48,089 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 18:32:12,253 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5483, 2.3487, 2.1083, 2.4852, 1.6582, 2.4780, 2.4035, 2.4539], + device='cuda:1'), covar=tensor([0.0122, 0.0528, 0.2620, 0.0091, 0.7131, 0.0082, 0.0649, 0.0044], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0148, 0.0251, 0.0093, 0.0239, 0.0095, 0.0160, 0.0099], + device='cuda:1'), out_proj_covar=tensor([7.7534e-05, 1.2281e-04, 1.9051e-04, 7.9875e-05, 1.9080e-04, 7.7100e-05, + 1.3110e-04, 8.1719e-05], device='cuda:1') +2023-03-20 18:32:14,286 INFO [train.py:935] (1/2) Epoch 4, validation: loss=0.2099, simple_loss=0.2916, pruned_loss=0.06411, over 1622729.00 frames. +2023-03-20 18:32:14,287 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-20 18:32:16,194 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 +2023-03-20 18:32:18,719 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.07 vs. limit=2.0 +2023-03-20 18:32:21,409 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 18:32:25,345 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5964, 4.8859, 5.0284, 5.0358, 4.8605, 4.6014, 5.2084, 4.9598], + device='cuda:1'), covar=tensor([0.0394, 0.0264, 0.0430, 0.0419, 0.0471, 0.0278, 0.0248, 0.0539], + device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0063, 0.0084, 0.0066, 0.0059, 0.0072, 0.0066, 0.0062], + device='cuda:1'), out_proj_covar=tensor([1.1426e-04, 8.8601e-05, 1.2662e-04, 1.0237e-04, 8.7446e-05, 1.0431e-04, + 9.5430e-05, 9.1542e-05], device='cuda:1') +2023-03-20 18:32:29,774 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.888e+02 3.865e+02 4.722e+02 6.124e+02 1.429e+03, threshold=9.445e+02, percent-clipped=4.0 +2023-03-20 18:32:30,896 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=8506.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:32:32,823 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 18:32:39,370 INFO [train.py:901] (1/2) Epoch 4, batch 50, loss[loss=0.312, simple_loss=0.3452, pruned_loss=0.1394, over 7328.00 frames. ], tot_loss[loss=0.2559, simple_loss=0.3027, pruned_loss=0.1046, over 322347.93 frames. ], batch size: 75, lr: 3.36e-02, grad_scale: 16.0 +2023-03-20 18:32:41,018 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 18:32:43,480 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 18:32:46,046 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 18:32:52,473 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.00 vs. limit=2.0 +2023-03-20 18:32:56,845 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0920, 2.3220, 2.1221, 2.5723, 1.3226, 2.8479, 2.2848, 2.9516], + device='cuda:1'), covar=tensor([0.0082, 0.0643, 0.2837, 0.0082, 0.8121, 0.0090, 0.0747, 0.0099], + device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0145, 0.0251, 0.0093, 0.0244, 0.0095, 0.0163, 0.0098], + device='cuda:1'), out_proj_covar=tensor([7.6421e-05, 1.2104e-04, 1.9126e-04, 7.9929e-05, 1.9369e-04, 7.7232e-05, + 1.3406e-04, 8.1965e-05], device='cuda:1') +2023-03-20 18:33:03,320 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 18:33:03,839 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 18:33:05,777 INFO [train.py:901] (1/2) Epoch 4, batch 100, loss[loss=0.2524, simple_loss=0.3095, pruned_loss=0.09761, over 7363.00 frames. ], tot_loss[loss=0.257, simple_loss=0.3048, pruned_loss=0.1046, over 570560.01 frames. ], batch size: 73, lr: 3.36e-02, grad_scale: 16.0 +2023-03-20 18:33:09,338 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3406, 3.7718, 3.5196, 3.7489, 3.7764, 3.5897, 3.5835, 3.2326], + device='cuda:1'), covar=tensor([0.0134, 0.0138, 0.0119, 0.0117, 0.0094, 0.0113, 0.0097, 0.0163], + device='cuda:1'), in_proj_covar=tensor([0.0022, 0.0023, 0.0022, 0.0021, 0.0019, 0.0022, 0.0025, 0.0024], + device='cuda:1'), out_proj_covar=tensor([5.1740e-05, 5.9616e-05, 5.6816e-05, 4.9402e-05, 4.8284e-05, 5.2987e-05, + 6.4103e-05, 5.8321e-05], device='cuda:1') +2023-03-20 18:33:21,690 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.466e+02 3.550e+02 4.535e+02 5.650e+02 1.214e+03, threshold=9.071e+02, percent-clipped=4.0 +2023-03-20 18:33:30,970 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8621.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:33:31,849 INFO [train.py:901] (1/2) Epoch 4, batch 150, loss[loss=0.2536, simple_loss=0.2933, pruned_loss=0.107, over 7228.00 frames. ], tot_loss[loss=0.2586, simple_loss=0.3059, pruned_loss=0.1056, over 765251.61 frames. ], batch size: 39, lr: 3.35e-02, grad_scale: 16.0 +2023-03-20 18:33:38,650 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.83 vs. limit=2.0 +2023-03-20 18:33:57,543 INFO [train.py:901] (1/2) Epoch 4, batch 200, loss[loss=0.2205, simple_loss=0.2677, pruned_loss=0.08669, over 6958.00 frames. ], tot_loss[loss=0.2586, simple_loss=0.3052, pruned_loss=0.106, over 911717.69 frames. ], batch size: 35, lr: 3.34e-02, grad_scale: 16.0 +2023-03-20 18:34:04,278 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 18:34:06,932 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.06 vs. limit=2.0 +2023-03-20 18:34:09,190 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 18:34:10,322 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6072, 1.8681, 1.1709, 2.2378, 0.8922, 1.9658, 1.6350, 1.1708], + device='cuda:1'), covar=tensor([0.0343, 0.0197, 0.0294, 0.0181, 0.0531, 0.0471, 0.0262, 0.0695], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0022, 0.0022, 0.0023, 0.0022, 0.0023, 0.0021, 0.0022], + device='cuda:1'), out_proj_covar=tensor([3.7050e-05, 3.5121e-05, 3.5627e-05, 3.7383e-05, 3.7474e-05, 4.0003e-05, + 3.5222e-05, 4.0919e-05], device='cuda:1') +2023-03-20 18:34:13,662 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.118e+02 3.602e+02 4.396e+02 5.497e+02 1.117e+03, threshold=8.791e+02, percent-clipped=2.0 +2023-03-20 18:34:14,695 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 18:34:23,230 INFO [train.py:901] (1/2) Epoch 4, batch 250, loss[loss=0.2716, simple_loss=0.3253, pruned_loss=0.109, over 7264.00 frames. ], tot_loss[loss=0.2608, simple_loss=0.3075, pruned_loss=0.107, over 1029153.65 frames. ], batch size: 64, lr: 3.33e-02, grad_scale: 16.0 +2023-03-20 18:34:27,284 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 18:34:42,552 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8760.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:34:44,809 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 +2023-03-20 18:34:48,060 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 18:34:49,072 INFO [train.py:901] (1/2) Epoch 4, batch 300, loss[loss=0.2713, simple_loss=0.319, pruned_loss=0.1118, over 7276.00 frames. ], tot_loss[loss=0.2602, simple_loss=0.3068, pruned_loss=0.1068, over 1118315.46 frames. ], batch size: 77, lr: 3.33e-02, grad_scale: 16.0 +2023-03-20 18:34:56,175 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7350, 4.0752, 3.9647, 4.0241, 3.6517, 4.0446, 4.4215, 4.3807], + device='cuda:1'), covar=tensor([0.0356, 0.0207, 0.0318, 0.0234, 0.0540, 0.0309, 0.0361, 0.0295], + device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0072, 0.0069, 0.0084, 0.0076, 0.0054, 0.0058, 0.0059], + device='cuda:1'), out_proj_covar=tensor([1.2716e-04, 1.2763e-04, 1.2004e-04, 1.5796e-04, 1.4541e-04, 9.7761e-05, + 1.0353e-04, 1.0557e-04], device='cuda:1') +2023-03-20 18:34:57,845 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 +2023-03-20 18:34:58,072 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 18:34:58,146 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1271, 4.5166, 4.3402, 4.5997, 4.3523, 4.0590, 4.7345, 4.5263], + device='cuda:1'), covar=tensor([0.0439, 0.0257, 0.0558, 0.0484, 0.0457, 0.0335, 0.0224, 0.0474], + device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0068, 0.0090, 0.0071, 0.0064, 0.0077, 0.0071, 0.0064], + device='cuda:1'), out_proj_covar=tensor([1.2441e-04, 9.6359e-05, 1.3810e-04, 1.1085e-04, 9.4980e-05, 1.1143e-04, + 1.0481e-04, 9.5207e-05], device='cuda:1') +2023-03-20 18:35:05,460 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.454e+02 3.664e+02 4.389e+02 5.371e+02 1.280e+03, threshold=8.778e+02, percent-clipped=5.0 +2023-03-20 18:35:14,075 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8821.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:35:14,943 INFO [train.py:901] (1/2) Epoch 4, batch 350, loss[loss=0.2213, simple_loss=0.2677, pruned_loss=0.0874, over 6977.00 frames. ], tot_loss[loss=0.2587, simple_loss=0.3055, pruned_loss=0.1059, over 1190726.68 frames. ], batch size: 35, lr: 3.32e-02, grad_scale: 16.0 +2023-03-20 18:35:32,166 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 18:35:41,195 INFO [train.py:901] (1/2) Epoch 4, batch 400, loss[loss=0.293, simple_loss=0.3302, pruned_loss=0.1279, over 7314.00 frames. ], tot_loss[loss=0.2581, simple_loss=0.3053, pruned_loss=0.1054, over 1246638.82 frames. ], batch size: 59, lr: 3.31e-02, grad_scale: 16.0 +2023-03-20 18:35:43,107 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 +2023-03-20 18:35:56,943 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.279e+02 3.727e+02 4.768e+02 5.740e+02 1.240e+03, threshold=9.535e+02, percent-clipped=4.0 +2023-03-20 18:36:06,305 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8921.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:36:07,224 INFO [train.py:901] (1/2) Epoch 4, batch 450, loss[loss=0.2707, simple_loss=0.3116, pruned_loss=0.1148, over 7267.00 frames. ], tot_loss[loss=0.2562, simple_loss=0.304, pruned_loss=0.1042, over 1292038.52 frames. ], batch size: 57, lr: 3.31e-02, grad_scale: 16.0 +2023-03-20 18:36:09,369 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8927.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:36:14,838 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 18:36:15,365 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 18:36:27,539 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1724, 1.9731, 1.0419, 2.0467, 1.0515, 1.8464, 1.2459, 1.2582], + device='cuda:1'), covar=tensor([0.0645, 0.0323, 0.0538, 0.0126, 0.0759, 0.0471, 0.0428, 0.0653], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0022, 0.0021, 0.0020, 0.0021, 0.0022, 0.0021, 0.0021], + device='cuda:1'), out_proj_covar=tensor([3.7504e-05, 3.5630e-05, 3.5206e-05, 3.3338e-05, 3.6442e-05, 3.9494e-05, + 3.3697e-05, 3.8915e-05], device='cuda:1') +2023-03-20 18:36:29,092 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.7561, 1.1208, 1.2009, 0.5785, 0.7357, 1.0162, 0.7540, 1.0434], + device='cuda:1'), covar=tensor([0.0805, 0.0806, 0.0192, 0.0622, 0.0893, 0.1557, 0.0701, 0.0955], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0031, 0.0025, 0.0028, 0.0024, 0.0029, 0.0031, 0.0029], + device='cuda:1'), out_proj_covar=tensor([3.9118e-05, 4.9121e-05, 3.4909e-05, 3.8209e-05, 3.6340e-05, 4.2076e-05, + 3.7694e-05, 3.9542e-05], device='cuda:1') +2023-03-20 18:36:31,033 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=8969.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:36:32,988 INFO [train.py:901] (1/2) Epoch 4, batch 500, loss[loss=0.2303, simple_loss=0.2887, pruned_loss=0.08591, over 7289.00 frames. ], tot_loss[loss=0.2564, simple_loss=0.3043, pruned_loss=0.1043, over 1326517.37 frames. ], batch size: 77, lr: 3.30e-02, grad_scale: 16.0 +2023-03-20 18:36:39,680 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8986.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:36:40,684 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8988.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:36:46,838 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 18:36:48,601 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 18:36:49,141 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 18:36:49,619 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.254e+02 3.528e+02 4.568e+02 6.123e+02 9.889e+02, threshold=9.136e+02, percent-clipped=2.0 +2023-03-20 18:36:51,158 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 18:36:53,411 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0399, 4.0097, 3.5609, 4.0391, 3.0216, 2.8008, 4.2441, 3.5973], + device='cuda:1'), covar=tensor([0.0064, 0.0038, 0.0125, 0.0039, 0.0194, 0.0470, 0.0104, 0.0343], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0083, 0.0134, 0.0092, 0.0163, 0.0229, 0.0102, 0.0209], + device='cuda:1'), out_proj_covar=tensor([7.7722e-05, 6.6140e-05, 9.8599e-05, 6.5588e-05, 1.2480e-04, 1.7414e-04, + 8.4678e-05, 1.6401e-04], device='cuda:1') +2023-03-20 18:36:54,425 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9013.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:36:55,280 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 18:36:59,323 INFO [train.py:901] (1/2) Epoch 4, batch 550, loss[loss=0.2442, simple_loss=0.2774, pruned_loss=0.1054, over 6933.00 frames. ], tot_loss[loss=0.2565, simple_loss=0.3045, pruned_loss=0.1043, over 1353522.73 frames. ], batch size: 35, lr: 3.29e-02, grad_scale: 16.0 +2023-03-20 18:37:02,648 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-03-20 18:37:05,994 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 18:37:11,943 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9047.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:37:14,843 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 18:37:18,526 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 18:37:25,073 INFO [train.py:901] (1/2) Epoch 4, batch 600, loss[loss=0.2646, simple_loss=0.3088, pruned_loss=0.1102, over 7289.00 frames. ], tot_loss[loss=0.2549, simple_loss=0.3034, pruned_loss=0.1032, over 1374979.85 frames. ], batch size: 57, lr: 3.29e-02, grad_scale: 16.0 +2023-03-20 18:37:25,743 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9074.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:37:26,128 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 18:37:41,223 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.856e+02 3.802e+02 4.609e+02 5.829e+02 9.078e+02, threshold=9.219e+02, percent-clipped=0.0 +2023-03-20 18:37:42,799 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 18:37:48,021 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9116.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:37:51,477 INFO [train.py:901] (1/2) Epoch 4, batch 650, loss[loss=0.2438, simple_loss=0.2967, pruned_loss=0.09548, over 7271.00 frames. ], tot_loss[loss=0.2545, simple_loss=0.303, pruned_loss=0.103, over 1389896.53 frames. ], batch size: 47, lr: 3.28e-02, grad_scale: 16.0 +2023-03-20 18:37:52,501 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 18:37:59,107 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4022, 3.6830, 3.9148, 3.7612, 4.0518, 4.0749, 4.3651, 3.9452], + device='cuda:1'), covar=tensor([0.0104, 0.0163, 0.0187, 0.0175, 0.0162, 0.0130, 0.0191, 0.0160], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0034, 0.0038, 0.0031, 0.0043, 0.0040, 0.0033, 0.0033], + device='cuda:1'), out_proj_covar=tensor([6.1949e-05, 7.2247e-05, 7.8823e-05, 6.6292e-05, 9.5385e-05, 8.5600e-05, + 7.7729e-05, 6.5738e-05], device='cuda:1') +2023-03-20 18:38:08,917 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 18:38:17,009 INFO [train.py:901] (1/2) Epoch 4, batch 700, loss[loss=0.1878, simple_loss=0.2424, pruned_loss=0.06666, over 6947.00 frames. ], tot_loss[loss=0.2542, simple_loss=0.3027, pruned_loss=0.1029, over 1401161.89 frames. ], batch size: 35, lr: 3.27e-02, grad_scale: 16.0 +2023-03-20 18:38:17,530 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 18:38:24,741 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.05 vs. limit=2.0 +2023-03-20 18:38:33,364 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.850e+02 3.888e+02 4.537e+02 6.098e+02 1.462e+03, threshold=9.073e+02, percent-clipped=2.0 +2023-03-20 18:38:41,445 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 18:38:42,386 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 18:38:42,905 INFO [train.py:901] (1/2) Epoch 4, batch 750, loss[loss=0.2053, simple_loss=0.2537, pruned_loss=0.07846, over 7032.00 frames. ], tot_loss[loss=0.2549, simple_loss=0.3032, pruned_loss=0.1033, over 1410726.99 frames. ], batch size: 35, lr: 3.26e-02, grad_scale: 16.0 +2023-03-20 18:38:55,640 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 18:38:57,887 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0001, 1.6784, 1.9493, 1.8646, 1.9776, 2.0335, 2.3282, 2.0358], + device='cuda:1'), covar=tensor([0.0633, 0.1051, 0.1114, 0.0613, 0.0818, 0.0830, 0.0545, 0.0891], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0034, 0.0034, 0.0036, 0.0034, 0.0038, 0.0035, 0.0035], + device='cuda:1'), out_proj_covar=tensor([6.6500e-05, 5.5911e-05, 5.8178e-05, 5.9900e-05, 5.8492e-05, 6.3925e-05, + 6.2690e-05, 6.0550e-05], device='cuda:1') +2023-03-20 18:39:00,783 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 18:39:06,833 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 18:39:08,372 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 18:39:08,862 INFO [train.py:901] (1/2) Epoch 4, batch 800, loss[loss=0.2329, simple_loss=0.293, pruned_loss=0.08638, over 7365.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.304, pruned_loss=0.1034, over 1418629.56 frames. ], batch size: 63, lr: 3.26e-02, grad_scale: 16.0 +2023-03-20 18:39:14,013 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9283.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:39:16,468 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.90 vs. limit=2.0 +2023-03-20 18:39:19,086 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 18:39:25,238 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.196e+02 3.605e+02 4.493e+02 5.434e+02 9.732e+02, threshold=8.986e+02, percent-clipped=1.0 +2023-03-20 18:39:34,822 INFO [train.py:901] (1/2) Epoch 4, batch 850, loss[loss=0.203, simple_loss=0.2603, pruned_loss=0.0728, over 7237.00 frames. ], tot_loss[loss=0.2547, simple_loss=0.3033, pruned_loss=0.1031, over 1422228.51 frames. ], batch size: 45, lr: 3.25e-02, grad_scale: 16.0 +2023-03-20 18:39:37,372 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 18:39:37,381 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 18:39:43,534 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 18:39:45,163 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9342.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:39:47,156 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 18:39:51,231 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4748, 4.8296, 4.9099, 4.9436, 4.6687, 4.2683, 5.0921, 4.7748], + device='cuda:1'), covar=tensor([0.0545, 0.0307, 0.0523, 0.0503, 0.0472, 0.0356, 0.0324, 0.0540], + device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0070, 0.0093, 0.0073, 0.0065, 0.0079, 0.0073, 0.0066], + device='cuda:1'), out_proj_covar=tensor([1.3051e-04, 9.9973e-05, 1.4161e-04, 1.1519e-04, 9.6988e-05, 1.1359e-04, + 1.0745e-04, 9.9303e-05], device='cuda:1') +2023-03-20 18:39:58,741 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9369.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:40:00,730 INFO [train.py:901] (1/2) Epoch 4, batch 900, loss[loss=0.2634, simple_loss=0.3028, pruned_loss=0.112, over 7368.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3027, pruned_loss=0.1023, over 1427656.67 frames. ], batch size: 73, lr: 3.24e-02, grad_scale: 16.0 +2023-03-20 18:40:11,502 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-20 18:40:17,375 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.070e+02 3.493e+02 4.561e+02 5.943e+02 1.247e+03, threshold=9.122e+02, percent-clipped=7.0 +2023-03-20 18:40:22,598 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9414.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:40:23,584 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9416.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:40:24,523 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 18:40:27,092 INFO [train.py:901] (1/2) Epoch 4, batch 950, loss[loss=0.2912, simple_loss=0.3348, pruned_loss=0.1238, over 6760.00 frames. ], tot_loss[loss=0.2559, simple_loss=0.3047, pruned_loss=0.1035, over 1431000.66 frames. ], batch size: 106, lr: 3.24e-02, grad_scale: 16.0 +2023-03-20 18:40:49,158 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=9464.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:40:49,625 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 18:40:53,605 INFO [train.py:901] (1/2) Epoch 4, batch 1000, loss[loss=0.2262, simple_loss=0.2712, pruned_loss=0.09061, over 7270.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.3033, pruned_loss=0.1028, over 1433537.57 frames. ], batch size: 52, lr: 3.23e-02, grad_scale: 16.0 +2023-03-20 18:40:53,907 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 +2023-03-20 18:40:54,750 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9475.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:40:59,444 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 18:41:09,171 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.060e+02 3.756e+02 4.692e+02 5.891e+02 2.084e+03, threshold=9.385e+02, percent-clipped=6.0 +2023-03-20 18:41:09,334 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5468, 2.8209, 2.1250, 2.4715, 2.7132, 2.8031, 1.1900, 2.4999], + device='cuda:1'), covar=tensor([0.0481, 0.0172, 0.0977, 0.0987, 0.0491, 0.0768, 0.1684, 0.0492], + device='cuda:1'), in_proj_covar=tensor([0.0023, 0.0022, 0.0025, 0.0027, 0.0023, 0.0025, 0.0030, 0.0025], + device='cuda:1'), out_proj_covar=tensor([4.4117e-05, 3.5813e-05, 4.7610e-05, 4.9763e-05, 4.2219e-05, 4.6478e-05, + 5.5204e-05, 4.5803e-05], device='cuda:1') +2023-03-20 18:41:09,703 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 18:41:19,221 INFO [train.py:901] (1/2) Epoch 4, batch 1050, loss[loss=0.2613, simple_loss=0.3026, pruned_loss=0.11, over 7226.00 frames. ], tot_loss[loss=0.2543, simple_loss=0.3034, pruned_loss=0.1026, over 1435832.42 frames. ], batch size: 45, lr: 3.22e-02, grad_scale: 16.0 +2023-03-20 18:41:32,355 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 18:41:36,930 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 18:41:43,960 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1289, 3.5652, 3.0674, 3.3396, 3.4694, 3.3300, 3.1777, 2.9031], + device='cuda:1'), covar=tensor([0.0110, 0.0106, 0.0126, 0.0097, 0.0083, 0.0089, 0.0146, 0.0159], + device='cuda:1'), in_proj_covar=tensor([0.0023, 0.0023, 0.0023, 0.0021, 0.0019, 0.0023, 0.0027, 0.0026], + device='cuda:1'), out_proj_covar=tensor([5.8280e-05, 6.5194e-05, 6.6524e-05, 5.1952e-05, 5.2977e-05, 6.0331e-05, + 7.3351e-05, 6.7931e-05], device='cuda:1') +2023-03-20 18:41:44,815 INFO [train.py:901] (1/2) Epoch 4, batch 1100, loss[loss=0.2243, simple_loss=0.2826, pruned_loss=0.08303, over 7320.00 frames. ], tot_loss[loss=0.2532, simple_loss=0.3025, pruned_loss=0.1019, over 1437490.07 frames. ], batch size: 49, lr: 3.22e-02, grad_scale: 16.0 +2023-03-20 18:41:49,949 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9583.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:41:50,439 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9584.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:42:01,278 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.878e+02 3.483e+02 4.054e+02 5.375e+02 1.176e+03, threshold=8.109e+02, percent-clipped=1.0 +2023-03-20 18:42:05,398 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3494, 1.4455, 1.4642, 1.4542, 1.2037, 1.7845, 1.4755, 1.1526], + device='cuda:1'), covar=tensor([0.0784, 0.0465, 0.0270, 0.0727, 0.0667, 0.0549, 0.0382, 0.1014], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0023, 0.0021, 0.0021, 0.0021, 0.0022, 0.0021, 0.0022], + device='cuda:1'), out_proj_covar=tensor([3.8546e-05, 3.8697e-05, 3.5852e-05, 3.6412e-05, 3.8406e-05, 4.0098e-05, + 3.5356e-05, 4.1116e-05], device='cuda:1') +2023-03-20 18:42:05,789 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 18:42:05,801 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 18:42:10,812 INFO [train.py:901] (1/2) Epoch 4, batch 1150, loss[loss=0.2393, simple_loss=0.3017, pruned_loss=0.08844, over 7279.00 frames. ], tot_loss[loss=0.2521, simple_loss=0.3017, pruned_loss=0.1013, over 1440242.97 frames. ], batch size: 77, lr: 3.21e-02, grad_scale: 16.0 +2023-03-20 18:42:14,924 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=9631.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:42:15,161 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-03-20 18:42:19,004 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 18:42:19,497 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 18:42:21,086 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9642.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:42:22,663 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9645.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:42:29,065 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9658.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:42:31,582 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7681, 3.9278, 3.8836, 3.8801, 3.5027, 3.8551, 4.3236, 4.2898], + device='cuda:1'), covar=tensor([0.0286, 0.0188, 0.0254, 0.0235, 0.0555, 0.0282, 0.0250, 0.0175], + device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0074, 0.0065, 0.0087, 0.0074, 0.0057, 0.0060, 0.0058], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 18:42:32,655 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0217, 1.5410, 2.0175, 1.8820, 2.3208, 2.1441, 2.1619, 2.4741], + device='cuda:1'), covar=tensor([0.0386, 0.0890, 0.0685, 0.0502, 0.1229, 0.0674, 0.0741, 0.0809], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0035, 0.0033, 0.0035, 0.0033, 0.0037, 0.0034, 0.0035], + device='cuda:1'), out_proj_covar=tensor([7.0092e-05, 6.0487e-05, 5.8332e-05, 5.9786e-05, 5.9895e-05, 6.4049e-05, + 6.2410e-05, 6.2317e-05], device='cuda:1') +2023-03-20 18:42:34,629 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9669.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:42:36,526 INFO [train.py:901] (1/2) Epoch 4, batch 1200, loss[loss=0.2551, simple_loss=0.3078, pruned_loss=0.1012, over 7266.00 frames. ], tot_loss[loss=0.252, simple_loss=0.3018, pruned_loss=0.1011, over 1441326.32 frames. ], batch size: 57, lr: 3.20e-02, grad_scale: 16.0 +2023-03-20 18:42:45,941 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=9690.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:42:50,461 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 18:42:52,839 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.497e+02 3.682e+02 4.480e+02 5.873e+02 1.157e+03, threshold=8.961e+02, percent-clipped=5.0 +2023-03-20 18:42:59,411 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=9717.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:43:01,164 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9719.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:43:03,037 INFO [train.py:901] (1/2) Epoch 4, batch 1250, loss[loss=0.2698, simple_loss=0.3164, pruned_loss=0.1116, over 7247.00 frames. ], tot_loss[loss=0.2535, simple_loss=0.3027, pruned_loss=0.1021, over 1441085.13 frames. ], batch size: 93, lr: 3.20e-02, grad_scale: 16.0 +2023-03-20 18:43:13,402 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9743.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:43:14,278 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 18:43:18,201 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 18:43:19,184 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 18:43:26,787 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9770.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:43:28,147 INFO [train.py:901] (1/2) Epoch 4, batch 1300, loss[loss=0.2706, simple_loss=0.3133, pruned_loss=0.1139, over 7373.00 frames. ], tot_loss[loss=0.2531, simple_loss=0.3026, pruned_loss=0.1019, over 1443654.19 frames. ], batch size: 65, lr: 3.19e-02, grad_scale: 32.0 +2023-03-20 18:43:35,463 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2059, 1.9902, 2.3057, 2.6388, 2.6532, 2.9334, 2.6579, 2.4608], + device='cuda:1'), covar=tensor([0.1019, 0.0732, 0.1679, 0.0227, 0.0075, 0.0043, 0.0062, 0.0034], + device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0154, 0.0205, 0.0108, 0.0078, 0.0082, 0.0083, 0.0084], + device='cuda:1'), out_proj_covar=tensor([1.8457e-04, 1.3931e-04, 1.7494e-04, 1.0347e-04, 7.0424e-05, 7.4303e-05, + 7.6338e-05, 7.7238e-05], device='cuda:1') +2023-03-20 18:43:42,329 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 18:43:44,625 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.204e+02 3.860e+02 4.926e+02 5.872e+02 1.226e+03, threshold=9.851e+02, percent-clipped=3.0 +2023-03-20 18:43:45,312 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 18:43:45,447 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9804.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:43:49,295 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 18:43:50,715 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-03-20 18:43:51,034 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8675, 1.6416, 1.7548, 1.5067, 2.1297, 1.7993, 1.9440, 2.0124], + device='cuda:1'), covar=tensor([0.1145, 0.0814, 0.0832, 0.1445, 0.0719, 0.0586, 0.0753, 0.0751], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0034, 0.0035, 0.0037, 0.0034, 0.0037, 0.0036, 0.0036], + device='cuda:1'), out_proj_covar=tensor([7.6184e-05, 6.0567e-05, 6.2327e-05, 6.3772e-05, 6.2195e-05, 6.6088e-05, + 6.7398e-05, 6.4840e-05], device='cuda:1') +2023-03-20 18:43:54,833 INFO [train.py:901] (1/2) Epoch 4, batch 1350, loss[loss=0.2527, simple_loss=0.2997, pruned_loss=0.1028, over 7344.00 frames. ], tot_loss[loss=0.2529, simple_loss=0.3024, pruned_loss=0.1017, over 1442554.35 frames. ], batch size: 63, lr: 3.18e-02, grad_scale: 32.0 +2023-03-20 18:43:59,395 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 18:44:20,605 INFO [train.py:901] (1/2) Epoch 4, batch 1400, loss[loss=0.3118, simple_loss=0.3488, pruned_loss=0.1374, over 7218.00 frames. ], tot_loss[loss=0.2523, simple_loss=0.3021, pruned_loss=0.1013, over 1442969.56 frames. ], batch size: 93, lr: 3.18e-02, grad_scale: 32.0 +2023-03-20 18:44:31,214 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 18:44:37,216 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 3.895e+02 4.499e+02 5.738e+02 1.302e+03, threshold=8.998e+02, percent-clipped=3.0 +2023-03-20 18:44:37,587 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.10 vs. limit=5.0 +2023-03-20 18:44:40,830 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0005, 4.1760, 4.1422, 4.1905, 3.8009, 4.1643, 4.4035, 4.5131], + device='cuda:1'), covar=tensor([0.0277, 0.0201, 0.0210, 0.0186, 0.0491, 0.0183, 0.0446, 0.0260], + device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0078, 0.0071, 0.0090, 0.0077, 0.0061, 0.0063, 0.0061], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 18:44:46,413 INFO [train.py:901] (1/2) Epoch 4, batch 1450, loss[loss=0.2584, simple_loss=0.3113, pruned_loss=0.1028, over 7354.00 frames. ], tot_loss[loss=0.2527, simple_loss=0.3021, pruned_loss=0.1016, over 1443757.38 frames. ], batch size: 73, lr: 3.17e-02, grad_scale: 16.0 +2023-03-20 18:44:55,063 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9940.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:44:56,998 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 18:45:12,168 INFO [train.py:901] (1/2) Epoch 4, batch 1500, loss[loss=0.2444, simple_loss=0.2951, pruned_loss=0.09688, over 7322.00 frames. ], tot_loss[loss=0.2528, simple_loss=0.3023, pruned_loss=0.1017, over 1443245.80 frames. ], batch size: 59, lr: 3.17e-02, grad_scale: 16.0 +2023-03-20 18:45:13,716 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 18:45:29,013 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.227e+02 3.489e+02 4.498e+02 5.516e+02 1.678e+03, threshold=8.996e+02, percent-clipped=8.0 +2023-03-20 18:45:33,680 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10014.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:45:37,616 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 18:45:38,111 INFO [train.py:901] (1/2) Epoch 4, batch 1550, loss[loss=0.2908, simple_loss=0.3364, pruned_loss=0.1226, over 6721.00 frames. ], tot_loss[loss=0.2527, simple_loss=0.3019, pruned_loss=0.1018, over 1441082.05 frames. ], batch size: 106, lr: 3.16e-02, grad_scale: 16.0 +2023-03-20 18:45:40,314 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4434, 2.0617, 2.0217, 2.7564, 3.0169, 3.0666, 3.0055, 2.5873], + device='cuda:1'), covar=tensor([0.1131, 0.0689, 0.1884, 0.0252, 0.0078, 0.0061, 0.0110, 0.0058], + device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0155, 0.0210, 0.0115, 0.0084, 0.0086, 0.0084, 0.0088], + device='cuda:1'), out_proj_covar=tensor([1.9355e-04, 1.4055e-04, 1.8026e-04, 1.0984e-04, 7.7880e-05, 7.8945e-05, + 7.7729e-05, 8.1165e-05], device='cuda:1') +2023-03-20 18:45:46,324 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10039.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:45:58,497 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10062.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 18:46:02,648 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10070.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:46:04,053 INFO [train.py:901] (1/2) Epoch 4, batch 1600, loss[loss=0.2348, simple_loss=0.2925, pruned_loss=0.08857, over 7238.00 frames. ], tot_loss[loss=0.2503, simple_loss=0.2998, pruned_loss=0.1003, over 1442418.70 frames. ], batch size: 55, lr: 3.15e-02, grad_scale: 16.0 +2023-03-20 18:46:08,428 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 +2023-03-20 18:46:11,173 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 18:46:12,174 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 18:46:14,789 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 18:46:17,902 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10099.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:46:18,513 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10100.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 18:46:19,966 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8188, 1.6778, 1.3508, 1.1878, 1.3204, 2.1635, 1.4940, 1.5694], + device='cuda:1'), covar=tensor([0.0489, 0.0680, 0.0392, 0.0329, 0.0938, 0.0392, 0.0525, 0.0114], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0023, 0.0021, 0.0020, 0.0021, 0.0021, 0.0021, 0.0020], + device='cuda:1'), out_proj_covar=tensor([3.9132e-05, 4.0486e-05, 3.5887e-05, 3.4727e-05, 3.8924e-05, 4.0317e-05, + 3.6372e-05, 3.8719e-05], device='cuda:1') +2023-03-20 18:46:20,786 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.208e+02 3.857e+02 4.437e+02 5.410e+02 1.141e+03, threshold=8.873e+02, percent-clipped=2.0 +2023-03-20 18:46:24,255 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 18:46:27,349 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=10118.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:46:28,316 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 18:46:29,718 INFO [train.py:901] (1/2) Epoch 4, batch 1650, loss[loss=0.2356, simple_loss=0.2913, pruned_loss=0.08996, over 7334.00 frames. ], tot_loss[loss=0.2519, simple_loss=0.3015, pruned_loss=0.1011, over 1444086.75 frames. ], batch size: 54, lr: 3.15e-02, grad_scale: 16.0 +2023-03-20 18:46:29,883 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10123.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 18:46:36,421 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 18:46:53,711 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 18:46:56,335 INFO [train.py:901] (1/2) Epoch 4, batch 1700, loss[loss=0.1919, simple_loss=0.2253, pruned_loss=0.07923, over 5970.00 frames. ], tot_loss[loss=0.2514, simple_loss=0.3012, pruned_loss=0.1008, over 1442357.59 frames. ], batch size: 25, lr: 3.14e-02, grad_scale: 16.0 +2023-03-20 18:46:57,854 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 18:47:07,903 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 18:47:12,608 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.114e+02 3.691e+02 4.418e+02 5.552e+02 1.158e+03, threshold=8.837e+02, percent-clipped=2.0 +2023-03-20 18:47:19,233 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-03-20 18:47:22,342 INFO [train.py:901] (1/2) Epoch 4, batch 1750, loss[loss=0.2219, simple_loss=0.2772, pruned_loss=0.08327, over 7345.00 frames. ], tot_loss[loss=0.253, simple_loss=0.3025, pruned_loss=0.1018, over 1442288.06 frames. ], batch size: 51, lr: 3.13e-02, grad_scale: 16.0 +2023-03-20 18:47:31,047 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10240.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:47:34,012 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 18:47:35,043 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 18:47:48,131 INFO [train.py:901] (1/2) Epoch 4, batch 1800, loss[loss=0.2181, simple_loss=0.2786, pruned_loss=0.0788, over 7182.00 frames. ], tot_loss[loss=0.2516, simple_loss=0.3014, pruned_loss=0.1009, over 1441954.23 frames. ], batch size: 39, lr: 3.13e-02, grad_scale: 16.0 +2023-03-20 18:47:55,670 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 +2023-03-20 18:47:55,816 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=10288.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:47:56,301 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 18:48:02,886 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8410, 3.9118, 3.2728, 3.9492, 2.8225, 2.3794, 3.8915, 3.2683], + device='cuda:1'), covar=tensor([0.0100, 0.0036, 0.0126, 0.0024, 0.0207, 0.0493, 0.0098, 0.0383], + device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0098, 0.0158, 0.0103, 0.0193, 0.0243, 0.0124, 0.0238], + device='cuda:1'), out_proj_covar=tensor([9.5988e-05, 8.2614e-05, 1.2216e-04, 7.9165e-05, 1.5525e-04, 1.9117e-04, + 1.0634e-04, 1.9365e-04], device='cuda:1') +2023-03-20 18:48:04,846 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.576e+02 3.726e+02 4.686e+02 5.832e+02 1.098e+03, threshold=9.372e+02, percent-clipped=3.0 +2023-03-20 18:48:09,482 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10314.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:48:09,912 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 18:48:13,938 INFO [train.py:901] (1/2) Epoch 4, batch 1850, loss[loss=0.2692, simple_loss=0.3185, pruned_loss=0.11, over 7274.00 frames. ], tot_loss[loss=0.2511, simple_loss=0.301, pruned_loss=0.1006, over 1440607.92 frames. ], batch size: 77, lr: 3.12e-02, grad_scale: 16.0 +2023-03-20 18:48:16,137 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8890, 3.8964, 3.0353, 4.1279, 2.9205, 2.3858, 4.0392, 3.1881], + device='cuda:1'), covar=tensor([0.0117, 0.0046, 0.0259, 0.0034, 0.0240, 0.0532, 0.0121, 0.0548], + device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0099, 0.0159, 0.0102, 0.0193, 0.0241, 0.0124, 0.0235], + device='cuda:1'), out_proj_covar=tensor([9.5991e-05, 8.3251e-05, 1.2319e-04, 7.8375e-05, 1.5522e-04, 1.8981e-04, + 1.0661e-04, 1.9147e-04], device='cuda:1') +2023-03-20 18:48:19,573 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 18:48:23,501 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 +2023-03-20 18:48:34,296 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=10362.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:48:37,335 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 18:48:39,757 INFO [train.py:901] (1/2) Epoch 4, batch 1900, loss[loss=0.2563, simple_loss=0.3117, pruned_loss=0.1005, over 7332.00 frames. ], tot_loss[loss=0.2502, simple_loss=0.3001, pruned_loss=0.1001, over 1441456.25 frames. ], batch size: 61, lr: 3.12e-02, grad_scale: 16.0 +2023-03-20 18:48:51,629 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10395.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 18:48:53,690 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10399.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:48:56,777 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.239e+02 3.747e+02 4.648e+02 5.275e+02 9.127e+02, threshold=9.296e+02, percent-clipped=0.0 +2023-03-20 18:48:59,005 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10409.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:49:02,388 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 18:49:03,472 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10418.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 18:49:06,421 INFO [train.py:901] (1/2) Epoch 4, batch 1950, loss[loss=0.3251, simple_loss=0.3463, pruned_loss=0.1519, over 6695.00 frames. ], tot_loss[loss=0.2503, simple_loss=0.3003, pruned_loss=0.1001, over 1440841.84 frames. ], batch size: 107, lr: 3.11e-02, grad_scale: 16.0 +2023-03-20 18:49:14,679 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 18:49:18,754 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 18:49:18,798 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=10447.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:49:19,740 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 18:49:22,379 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10454.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:49:30,442 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10470.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:49:31,810 INFO [train.py:901] (1/2) Epoch 4, batch 2000, loss[loss=0.2426, simple_loss=0.3022, pruned_loss=0.09144, over 7324.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3005, pruned_loss=0.1003, over 1441307.93 frames. ], batch size: 83, lr: 3.10e-02, grad_scale: 16.0 +2023-03-20 18:49:35,704 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 18:49:40,707 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.00 vs. limit=2.0 +2023-03-20 18:49:47,140 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5751, 3.9418, 3.8379, 3.8306, 3.4251, 3.9014, 4.1905, 4.2026], + device='cuda:1'), covar=tensor([0.0399, 0.0219, 0.0263, 0.0259, 0.0497, 0.0280, 0.0323, 0.0230], + device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0081, 0.0076, 0.0096, 0.0084, 0.0063, 0.0068, 0.0065], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 18:49:48,118 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 18:49:48,607 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.129e+02 3.642e+02 4.667e+02 5.884e+02 9.942e+02, threshold=9.334e+02, percent-clipped=1.0 +2023-03-20 18:49:54,535 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10515.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 18:49:56,297 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 +2023-03-20 18:49:56,909 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 18:49:58,397 INFO [train.py:901] (1/2) Epoch 4, batch 2050, loss[loss=0.2614, simple_loss=0.3025, pruned_loss=0.1102, over 7318.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.2997, pruned_loss=0.0995, over 1442918.62 frames. ], batch size: 75, lr: 3.10e-02, grad_scale: 16.0 +2023-03-20 18:50:13,402 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 +2023-03-20 18:50:24,357 INFO [train.py:901] (1/2) Epoch 4, batch 2100, loss[loss=0.1943, simple_loss=0.2523, pruned_loss=0.06811, over 7340.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.2979, pruned_loss=0.09827, over 1440581.33 frames. ], batch size: 44, lr: 3.09e-02, grad_scale: 16.0 +2023-03-20 18:50:30,351 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 18:50:33,319 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 18:50:41,270 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.704e+02 3.659e+02 4.368e+02 5.269e+02 1.016e+03, threshold=8.736e+02, percent-clipped=1.0 +2023-03-20 18:50:43,449 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7847, 3.3283, 3.3926, 3.3477, 3.3616, 3.3920, 3.5803, 3.3475], + device='cuda:1'), covar=tensor([0.0117, 0.0260, 0.0219, 0.0322, 0.0299, 0.0231, 0.0254, 0.0280], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0042, 0.0044, 0.0039, 0.0052, 0.0048, 0.0042, 0.0039], + device='cuda:1'), out_proj_covar=tensor([7.9611e-05, 9.6901e-05, 9.8989e-05, 8.8865e-05, 1.2064e-04, 1.1374e-04, + 1.0381e-04, 8.6253e-05], device='cuda:1') +2023-03-20 18:50:50,267 INFO [train.py:901] (1/2) Epoch 4, batch 2150, loss[loss=0.2824, simple_loss=0.3278, pruned_loss=0.1185, over 7301.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.2987, pruned_loss=0.09876, over 1442099.43 frames. ], batch size: 59, lr: 3.09e-02, grad_scale: 16.0 +2023-03-20 18:51:16,239 INFO [train.py:901] (1/2) Epoch 4, batch 2200, loss[loss=0.2668, simple_loss=0.3207, pruned_loss=0.1065, over 7225.00 frames. ], tot_loss[loss=0.2496, simple_loss=0.3001, pruned_loss=0.09958, over 1443847.85 frames. ], batch size: 93, lr: 3.08e-02, grad_scale: 16.0 +2023-03-20 18:51:20,459 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 18:51:28,107 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10695.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:51:29,131 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3336, 1.7701, 1.6265, 1.7447, 2.2369, 1.9542, 1.7398, 2.3473], + device='cuda:1'), covar=tensor([0.0413, 0.0201, 0.1524, 0.1187, 0.0364, 0.0770, 0.0939, 0.0332], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0023, 0.0024, 0.0025, 0.0023, 0.0023, 0.0027, 0.0022], + device='cuda:1'), out_proj_covar=tensor([4.9140e-05, 4.2283e-05, 4.8911e-05, 5.0837e-05, 4.6684e-05, 4.5487e-05, + 5.3868e-05, 4.4096e-05], device='cuda:1') +2023-03-20 18:51:33,050 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.783e+02 3.779e+02 4.552e+02 5.953e+02 1.477e+03, threshold=9.104e+02, percent-clipped=10.0 +2023-03-20 18:51:39,718 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10718.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 18:51:42,119 INFO [train.py:901] (1/2) Epoch 4, batch 2250, loss[loss=0.2077, simple_loss=0.261, pruned_loss=0.07721, over 6959.00 frames. ], tot_loss[loss=0.2492, simple_loss=0.2996, pruned_loss=0.09938, over 1442735.47 frames. ], batch size: 35, lr: 3.07e-02, grad_scale: 16.0 +2023-03-20 18:51:52,469 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=10743.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:51:53,060 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10744.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:51:53,953 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 18:51:53,965 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 18:52:04,181 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10765.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:52:04,691 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=10766.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 18:52:08,282 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 18:52:08,772 INFO [train.py:901] (1/2) Epoch 4, batch 2300, loss[loss=0.2582, simple_loss=0.305, pruned_loss=0.1057, over 7359.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.2988, pruned_loss=0.09871, over 1442404.08 frames. ], batch size: 63, lr: 3.07e-02, grad_scale: 16.0 +2023-03-20 18:52:09,441 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10774.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:52:11,748 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 +2023-03-20 18:52:25,628 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.241e+02 3.504e+02 4.106e+02 5.041e+02 1.029e+03, threshold=8.211e+02, percent-clipped=2.0 +2023-03-20 18:52:25,840 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10805.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:52:28,323 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10810.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 18:52:34,787 INFO [train.py:901] (1/2) Epoch 4, batch 2350, loss[loss=0.2494, simple_loss=0.3074, pruned_loss=0.09572, over 7249.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.2979, pruned_loss=0.09789, over 1442904.39 frames. ], batch size: 55, lr: 3.06e-02, grad_scale: 16.0 +2023-03-20 18:52:41,577 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.8822, 0.9558, 1.4110, 0.8760, 0.6059, 0.6364, 0.7458, 0.7171], + device='cuda:1'), covar=tensor([0.0627, 0.0692, 0.0167, 0.0524, 0.1149, 0.1275, 0.0420, 0.0686], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0034, 0.0026, 0.0031, 0.0029, 0.0029, 0.0033, 0.0030], + device='cuda:1'), out_proj_covar=tensor([4.0796e-05, 5.8144e-05, 3.6257e-05, 4.6487e-05, 4.4215e-05, 4.7279e-05, + 4.7960e-05, 4.4965e-05], device='cuda:1') +2023-03-20 18:52:41,590 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10835.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:52:55,239 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 +2023-03-20 18:52:56,406 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 18:52:56,673 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 +2023-03-20 18:52:57,050 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.8572, 0.8303, 0.9670, 0.8781, 0.9374, 0.8593, 0.8208, 0.7376], + device='cuda:1'), covar=tensor([0.0173, 0.0210, 0.0371, 0.0374, 0.0188, 0.0262, 0.0273, 0.0345], + device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0020, 0.0017, 0.0019, 0.0020, 0.0018, 0.0019, 0.0020], + device='cuda:1'), out_proj_covar=tensor([2.4527e-05, 2.2366e-05, 2.3062e-05, 2.4630e-05, 2.1935e-05, 2.3282e-05, + 2.5980e-05, 2.8008e-05], device='cuda:1') +2023-03-20 18:53:01,347 INFO [train.py:901] (1/2) Epoch 4, batch 2400, loss[loss=0.1984, simple_loss=0.2652, pruned_loss=0.06577, over 7359.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.2963, pruned_loss=0.09699, over 1442272.91 frames. ], batch size: 54, lr: 3.06e-02, grad_scale: 16.0 +2023-03-20 18:53:02,172 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 +2023-03-20 18:53:02,394 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 18:53:07,596 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10885.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 18:53:13,001 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 18:53:16,073 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 18:53:17,634 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.227e+02 3.878e+02 4.809e+02 5.497e+02 2.353e+03, threshold=9.618e+02, percent-clipped=8.0 +2023-03-20 18:53:27,297 INFO [train.py:901] (1/2) Epoch 4, batch 2450, loss[loss=0.2599, simple_loss=0.3124, pruned_loss=0.1037, over 7301.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.2965, pruned_loss=0.09713, over 1442878.02 frames. ], batch size: 86, lr: 3.05e-02, grad_scale: 16.0 +2023-03-20 18:53:39,769 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10946.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 18:53:43,128 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 18:53:48,198 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0331, 4.0756, 3.9323, 3.3632, 3.8299, 2.6949, 1.7215, 3.9871], + device='cuda:1'), covar=tensor([0.0034, 0.0123, 0.0078, 0.0087, 0.0029, 0.0458, 0.1195, 0.0061], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0054, 0.0043, 0.0042, 0.0067, 0.0085, 0.0044], + device='cuda:1'), out_proj_covar=tensor([5.1397e-05, 6.1709e-05, 7.8268e-05, 6.3570e-05, 5.3662e-05, 9.8724e-05, + 1.2511e-04, 6.1902e-05], device='cuda:1') +2023-03-20 18:53:53,060 INFO [train.py:901] (1/2) Epoch 4, batch 2500, loss[loss=0.2533, simple_loss=0.3076, pruned_loss=0.09952, over 7235.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.2966, pruned_loss=0.0973, over 1442476.74 frames. ], batch size: 93, lr: 3.04e-02, grad_scale: 16.0 +2023-03-20 18:54:03,946 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.14 vs. limit=2.0 +2023-03-20 18:54:07,990 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 18:54:09,458 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.005e+02 3.468e+02 4.610e+02 5.877e+02 1.140e+03, threshold=9.220e+02, percent-clipped=4.0 +2023-03-20 18:54:19,250 INFO [train.py:901] (1/2) Epoch 4, batch 2550, loss[loss=0.2421, simple_loss=0.2927, pruned_loss=0.0958, over 7304.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.2976, pruned_loss=0.09754, over 1443183.95 frames. ], batch size: 80, lr: 3.04e-02, grad_scale: 16.0 +2023-03-20 18:54:20,646 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-20 18:54:41,080 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11065.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:54:44,898 INFO [train.py:901] (1/2) Epoch 4, batch 2600, loss[loss=0.26, simple_loss=0.3037, pruned_loss=0.1082, over 7343.00 frames. ], tot_loss[loss=0.2474, simple_loss=0.2983, pruned_loss=0.0982, over 1441704.88 frames. ], batch size: 54, lr: 3.03e-02, grad_scale: 16.0 +2023-03-20 18:54:45,237 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 18:54:58,279 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11100.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:54:58,960 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 +2023-03-20 18:54:59,266 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5364, 3.7630, 3.2926, 3.5057, 3.7100, 3.6145, 3.2987, 3.3021], + device='cuda:1'), covar=tensor([0.0082, 0.0097, 0.0109, 0.0107, 0.0071, 0.0069, 0.0120, 0.0106], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0025, 0.0026, 0.0023, 0.0023, 0.0024, 0.0029, 0.0028], + device='cuda:1'), out_proj_covar=tensor([6.6452e-05, 7.6432e-05, 8.3714e-05, 6.5112e-05, 6.6464e-05, 6.9804e-05, + 8.7516e-05, 8.3208e-05], device='cuda:1') +2023-03-20 18:55:00,593 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.035e+02 3.420e+02 4.169e+02 5.328e+02 1.067e+03, threshold=8.339e+02, percent-clipped=1.0 +2023-03-20 18:55:02,689 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.8743, 0.8956, 1.3280, 1.2029, 0.9667, 0.5390, 0.8266, 0.8945], + device='cuda:1'), covar=tensor([0.0848, 0.1112, 0.0342, 0.0463, 0.0767, 0.1311, 0.0595, 0.0546], + device='cuda:1'), in_proj_covar=tensor([0.0026, 0.0032, 0.0025, 0.0028, 0.0024, 0.0026, 0.0030, 0.0029], + device='cuda:1'), out_proj_covar=tensor([3.8415e-05, 5.5601e-05, 3.4808e-05, 4.2712e-05, 3.8934e-05, 4.2978e-05, + 4.4016e-05, 4.2691e-05], device='cuda:1') +2023-03-20 18:55:03,125 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11110.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 18:55:04,610 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=11113.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:55:09,435 INFO [train.py:901] (1/2) Epoch 4, batch 2650, loss[loss=0.2033, simple_loss=0.2569, pruned_loss=0.07483, over 7187.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.2977, pruned_loss=0.09745, over 1443518.31 frames. ], batch size: 39, lr: 3.03e-02, grad_scale: 16.0 +2023-03-20 18:55:12,920 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11130.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:55:22,551 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.16 vs. limit=2.0 +2023-03-20 18:55:26,781 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=11158.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:55:34,659 INFO [train.py:901] (1/2) Epoch 4, batch 2700, loss[loss=0.2355, simple_loss=0.2896, pruned_loss=0.09072, over 7232.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.2978, pruned_loss=0.09734, over 1442723.26 frames. ], batch size: 93, lr: 3.02e-02, grad_scale: 16.0 +2023-03-20 18:55:35,283 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11174.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:55:50,731 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.463e+02 3.424e+02 4.477e+02 5.866e+02 1.423e+03, threshold=8.953e+02, percent-clipped=8.0 +2023-03-20 18:55:52,187 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-03-20 18:55:54,664 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 +2023-03-20 18:56:00,389 INFO [train.py:901] (1/2) Epoch 4, batch 2750, loss[loss=0.2448, simple_loss=0.2974, pruned_loss=0.09607, over 7357.00 frames. ], tot_loss[loss=0.2446, simple_loss=0.2961, pruned_loss=0.09657, over 1441245.62 frames. ], batch size: 63, lr: 3.02e-02, grad_scale: 16.0 +2023-03-20 18:56:06,640 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11235.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:56:09,622 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11241.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 18:56:09,872 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 +2023-03-20 18:56:25,296 INFO [train.py:901] (1/2) Epoch 4, batch 2800, loss[loss=0.2844, simple_loss=0.3317, pruned_loss=0.1186, over 7156.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.2962, pruned_loss=0.0964, over 1442007.14 frames. ], batch size: 98, lr: 3.01e-02, grad_scale: 16.0 +2023-03-20 18:56:50,511 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 18:56:50,564 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 18:56:50,589 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 18:56:50,605 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 18:56:50,775 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 18:56:50,791 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 18:56:50,801 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 18:56:50,897 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 18:56:50,971 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 18:56:50,987 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 18:56:51,003 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 18:56:51,186 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 18:56:51,191 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 18:56:51,198 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 18:56:51,670 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 18:56:51,673 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 18:56:51,729 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 18:56:58,649 INFO [train.py:901] (1/2) Epoch 5, batch 0, loss[loss=0.244, simple_loss=0.3002, pruned_loss=0.09388, over 7363.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3002, pruned_loss=0.09388, over 7363.00 frames. ], batch size: 73, lr: 2.90e-02, grad_scale: 16.0 +2023-03-20 18:56:58,649 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 18:57:14,516 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3096, 1.1409, 1.2892, 1.5607, 1.2485, 0.4972, 0.8846, 1.5914], + device='cuda:1'), covar=tensor([0.0529, 0.1213, 0.0199, 0.0632, 0.0959, 0.2089, 0.0323, 0.0386], + device='cuda:1'), in_proj_covar=tensor([0.0026, 0.0031, 0.0024, 0.0027, 0.0024, 0.0026, 0.0028, 0.0028], + device='cuda:1'), out_proj_covar=tensor([3.7796e-05, 5.3574e-05, 3.3089e-05, 4.1911e-05, 3.8912e-05, 4.2765e-05, + 4.1202e-05, 4.1520e-05], device='cuda:1') +2023-03-20 18:57:20,490 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7500, 3.7710, 3.7434, 3.2339, 3.6074, 2.6382, 1.6543, 4.1971], + device='cuda:1'), covar=tensor([0.0017, 0.0117, 0.0071, 0.0149, 0.0026, 0.0502, 0.1226, 0.0053], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0042, 0.0054, 0.0044, 0.0043, 0.0066, 0.0087, 0.0047], + device='cuda:1'), out_proj_covar=tensor([5.0941e-05, 6.6504e-05, 7.8384e-05, 6.4523e-05, 5.3866e-05, 9.7078e-05, + 1.2793e-04, 6.5185e-05], device='cuda:1') +2023-03-20 18:57:22,165 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7682, 2.1672, 2.3740, 2.4338, 1.7332, 2.7252, 2.4084, 2.8023], + device='cuda:1'), covar=tensor([0.0033, 0.0639, 0.1682, 0.0029, 0.4952, 0.0064, 0.0683, 0.0031], + device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0181, 0.0280, 0.0109, 0.0274, 0.0107, 0.0199, 0.0110], + device='cuda:1'), out_proj_covar=tensor([9.3705e-05, 1.5703e-04, 2.2274e-04, 9.5719e-05, 2.2956e-04, 9.6580e-05, + 1.6927e-04, 9.9041e-05], device='cuda:1') +2023-03-20 18:57:24,430 INFO [train.py:935] (1/2) Epoch 5, validation: loss=0.199, simple_loss=0.2825, pruned_loss=0.05778, over 1622729.00 frames. +2023-03-20 18:57:24,431 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-20 18:57:28,122 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3067, 2.4496, 2.2898, 3.1340, 2.8975, 3.2379, 2.8474, 2.6917], + device='cuda:1'), covar=tensor([0.0947, 0.0400, 0.1225, 0.0176, 0.0020, 0.0023, 0.0023, 0.0033], + device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0167, 0.0223, 0.0124, 0.0088, 0.0089, 0.0086, 0.0088], + device='cuda:1'), out_proj_covar=tensor([2.0659e-04, 1.5584e-04, 1.9656e-04, 1.2189e-04, 8.4582e-05, 8.3713e-05, + 8.4104e-05, 8.5314e-05], device='cuda:1') +2023-03-20 18:57:28,422 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.740e+02 3.320e+02 4.022e+02 5.081e+02 1.321e+03, threshold=8.045e+02, percent-clipped=4.0 +2023-03-20 18:57:30,540 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6514, 3.6600, 3.6661, 3.6439, 3.3409, 3.6516, 3.9881, 4.0311], + device='cuda:1'), covar=tensor([0.0257, 0.0206, 0.0268, 0.0260, 0.0459, 0.0363, 0.0285, 0.0205], + device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0078, 0.0074, 0.0092, 0.0084, 0.0062, 0.0065, 0.0063], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 18:57:30,960 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 18:57:41,990 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 18:57:48,521 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 18:57:49,495 INFO [train.py:901] (1/2) Epoch 5, batch 50, loss[loss=0.25, simple_loss=0.2998, pruned_loss=0.1, over 7304.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.2971, pruned_loss=0.09729, over 325858.18 frames. ], batch size: 86, lr: 2.90e-02, grad_scale: 16.0 +2023-03-20 18:57:50,533 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 18:57:53,094 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3730, 4.8129, 4.7408, 4.6851, 4.5993, 4.3648, 4.8343, 4.6023], + device='cuda:1'), covar=tensor([0.0429, 0.0263, 0.0517, 0.0492, 0.0287, 0.0359, 0.0370, 0.0447], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0079, 0.0094, 0.0077, 0.0067, 0.0087, 0.0078, 0.0070], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 18:57:53,524 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 18:57:55,850 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 +2023-03-20 18:58:03,974 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 +2023-03-20 18:58:10,836 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 18:58:11,376 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 18:58:11,443 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0367, 4.5638, 4.5634, 4.5029, 4.3793, 3.9322, 4.6037, 4.3823], + device='cuda:1'), covar=tensor([0.0637, 0.0351, 0.0585, 0.0604, 0.0477, 0.0433, 0.0511, 0.0758], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0080, 0.0096, 0.0079, 0.0070, 0.0089, 0.0081, 0.0072], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 18:58:15,836 INFO [train.py:901] (1/2) Epoch 5, batch 100, loss[loss=0.2682, simple_loss=0.3172, pruned_loss=0.1096, over 7287.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.298, pruned_loss=0.09664, over 573171.95 frames. ], batch size: 86, lr: 2.89e-02, grad_scale: 16.0 +2023-03-20 18:58:17,813 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11400.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:58:20,162 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.320e+02 3.482e+02 4.365e+02 5.725e+02 1.315e+03, threshold=8.730e+02, percent-clipped=7.0 +2023-03-20 18:58:32,664 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11430.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:58:32,716 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2860, 2.2824, 2.2731, 3.0562, 2.6338, 3.1055, 2.7634, 2.6096], + device='cuda:1'), covar=tensor([0.0989, 0.0500, 0.1346, 0.0203, 0.0038, 0.0042, 0.0027, 0.0031], + device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0167, 0.0217, 0.0125, 0.0089, 0.0089, 0.0083, 0.0087], + device='cuda:1'), out_proj_covar=tensor([2.0666e-04, 1.5598e-04, 1.9203e-04, 1.2151e-04, 8.4924e-05, 8.3147e-05, + 8.1806e-05, 8.3820e-05], device='cuda:1') +2023-03-20 18:58:41,106 INFO [train.py:901] (1/2) Epoch 5, batch 150, loss[loss=0.1612, simple_loss=0.2225, pruned_loss=0.04995, over 7054.00 frames. ], tot_loss[loss=0.243, simple_loss=0.2955, pruned_loss=0.09523, over 767450.27 frames. ], batch size: 35, lr: 2.89e-02, grad_scale: 16.0 +2023-03-20 18:58:41,665 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=11448.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:58:58,018 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=11478.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:59:07,537 INFO [train.py:901] (1/2) Epoch 5, batch 200, loss[loss=0.2472, simple_loss=0.298, pruned_loss=0.09817, over 7367.00 frames. ], tot_loss[loss=0.241, simple_loss=0.2934, pruned_loss=0.09434, over 916113.81 frames. ], batch size: 73, lr: 2.88e-02, grad_scale: 16.0 +2023-03-20 18:59:11,559 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.950e+02 3.488e+02 4.378e+02 5.184e+02 9.944e+02, threshold=8.756e+02, percent-clipped=2.0 +2023-03-20 18:59:12,065 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 18:59:16,510 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 18:59:23,560 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 18:59:24,127 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11530.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:59:25,211 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11532.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:59:29,982 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.52 vs. limit=2.0 +2023-03-20 18:59:30,290 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11541.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 18:59:33,187 INFO [train.py:901] (1/2) Epoch 5, batch 250, loss[loss=0.2271, simple_loss=0.2878, pruned_loss=0.08321, over 7312.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.2927, pruned_loss=0.09405, over 1033127.21 frames. ], batch size: 83, lr: 2.87e-02, grad_scale: 16.0 +2023-03-20 18:59:36,248 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 18:59:54,822 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=11589.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 18:59:56,901 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11593.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 18:59:57,256 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 18:59:58,663 INFO [train.py:901] (1/2) Epoch 5, batch 300, loss[loss=0.2709, simple_loss=0.3251, pruned_loss=0.1084, over 7199.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.2929, pruned_loss=0.09401, over 1126669.72 frames. ], batch size: 99, lr: 2.87e-02, grad_scale: 16.0 +2023-03-20 19:00:03,000 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.958e+02 3.432e+02 4.391e+02 5.389e+02 9.062e+02, threshold=8.782e+02, percent-clipped=1.0 +2023-03-20 19:00:06,022 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 19:00:18,640 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11635.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:00:21,367 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 +2023-03-20 19:00:25,219 INFO [train.py:901] (1/2) Epoch 5, batch 350, loss[loss=0.2587, simple_loss=0.299, pruned_loss=0.1092, over 7269.00 frames. ], tot_loss[loss=0.241, simple_loss=0.2931, pruned_loss=0.09442, over 1196460.59 frames. ], batch size: 52, lr: 2.86e-02, grad_scale: 16.0 +2023-03-20 19:00:29,098 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-03-20 19:00:41,251 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 19:00:48,296 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-03-20 19:00:49,992 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11696.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:00:50,326 INFO [train.py:901] (1/2) Epoch 5, batch 400, loss[loss=0.2582, simple_loss=0.3095, pruned_loss=0.1035, over 7275.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.2921, pruned_loss=0.09356, over 1250469.22 frames. ], batch size: 57, lr: 2.86e-02, grad_scale: 16.0 +2023-03-20 19:00:50,634 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.07 vs. limit=2.0 +2023-03-20 19:00:51,473 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11699.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:00:54,169 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.098e+02 3.501e+02 4.253e+02 5.454e+02 9.095e+02, threshold=8.507e+02, percent-clipped=1.0 +2023-03-20 19:01:16,686 INFO [train.py:901] (1/2) Epoch 5, batch 450, loss[loss=0.2039, simple_loss=0.2657, pruned_loss=0.07104, over 7297.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.2916, pruned_loss=0.09345, over 1291909.36 frames. ], batch size: 42, lr: 2.85e-02, grad_scale: 16.0 +2023-03-20 19:01:22,763 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 19:01:22,776 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 19:01:23,410 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11760.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:01:41,720 INFO [train.py:901] (1/2) Epoch 5, batch 500, loss[loss=0.2142, simple_loss=0.2739, pruned_loss=0.07724, over 7272.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.2916, pruned_loss=0.09306, over 1325774.66 frames. ], batch size: 47, lr: 2.85e-02, grad_scale: 16.0 +2023-03-20 19:01:46,704 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.845e+02 3.462e+02 4.263e+02 5.633e+02 1.413e+03, threshold=8.525e+02, percent-clipped=5.0 +2023-03-20 19:01:52,841 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11816.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:01:54,722 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 19:01:56,257 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 19:01:56,735 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 19:01:59,285 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 19:01:59,882 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3463, 3.8378, 3.9607, 3.9510, 3.6443, 4.0636, 4.2370, 3.7375], + device='cuda:1'), covar=tensor([0.0199, 0.0202, 0.0219, 0.0218, 0.0390, 0.0229, 0.0331, 0.0281], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0042, 0.0043, 0.0036, 0.0052, 0.0048, 0.0040, 0.0038], + device='cuda:1'), out_proj_covar=tensor([7.4397e-05, 9.9962e-05, 9.9675e-05, 8.7739e-05, 1.2619e-04, 1.1952e-04, + 1.0335e-04, 8.6174e-05], device='cuda:1') +2023-03-20 19:01:59,898 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11830.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:02:04,236 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 19:02:07,859 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11846.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:02:08,199 INFO [train.py:901] (1/2) Epoch 5, batch 550, loss[loss=0.2349, simple_loss=0.2891, pruned_loss=0.0904, over 7262.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.2922, pruned_loss=0.09321, over 1353116.96 frames. ], batch size: 52, lr: 2.84e-02, grad_scale: 16.0 +2023-03-20 19:02:14,289 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 19:02:22,366 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 19:02:22,461 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0534, 3.7249, 3.8268, 3.8500, 3.8132, 3.9230, 4.1105, 3.6646], + device='cuda:1'), covar=tensor([0.0239, 0.0172, 0.0158, 0.0167, 0.0191, 0.0138, 0.0234, 0.0180], + device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0042, 0.0043, 0.0036, 0.0051, 0.0047, 0.0040, 0.0037], + device='cuda:1'), out_proj_covar=tensor([7.1685e-05, 9.8753e-05, 1.0087e-04, 8.6677e-05, 1.2511e-04, 1.1618e-04, + 1.0170e-04, 8.3091e-05], device='cuda:1') +2023-03-20 19:02:23,521 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11877.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:02:23,929 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=11878.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:02:25,837 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 19:02:29,609 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11888.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:02:33,586 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 19:02:34,081 INFO [train.py:901] (1/2) Epoch 5, batch 600, loss[loss=0.1961, simple_loss=0.2502, pruned_loss=0.07104, over 6969.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.2917, pruned_loss=0.09238, over 1375043.84 frames. ], batch size: 35, lr: 2.84e-02, grad_scale: 32.0 +2023-03-20 19:02:38,649 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.069e+02 3.397e+02 4.178e+02 5.537e+02 1.097e+03, threshold=8.356e+02, percent-clipped=4.0 +2023-03-20 19:02:39,784 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11907.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:02:46,288 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11920.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:02:49,188 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 19:02:57,600 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 19:02:59,595 INFO [train.py:901] (1/2) Epoch 5, batch 650, loss[loss=0.2398, simple_loss=0.2918, pruned_loss=0.09394, over 7360.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.2925, pruned_loss=0.09261, over 1392253.42 frames. ], batch size: 73, lr: 2.83e-02, grad_scale: 32.0 +2023-03-20 19:03:00,237 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11948.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:03:16,858 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 19:03:17,488 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11981.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:03:18,921 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11984.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:03:22,868 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11991.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:03:25,766 INFO [train.py:901] (1/2) Epoch 5, batch 700, loss[loss=0.1753, simple_loss=0.2316, pruned_loss=0.05953, over 7180.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.2907, pruned_loss=0.09188, over 1402950.39 frames. ], batch size: 39, lr: 2.83e-02, grad_scale: 32.0 +2023-03-20 19:03:26,283 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 19:03:33,387 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.580e+02 3.642e+02 4.392e+02 5.273e+02 9.555e+02, threshold=8.784e+02, percent-clipped=4.0 +2023-03-20 19:03:35,583 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12009.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:03:39,287 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.07 vs. limit=2.0 +2023-03-20 19:03:52,630 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 19:03:53,063 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 19:03:53,726 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12045.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:03:54,568 INFO [train.py:901] (1/2) Epoch 5, batch 750, loss[loss=0.195, simple_loss=0.2538, pruned_loss=0.0681, over 7183.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.2903, pruned_loss=0.09164, over 1412570.06 frames. ], batch size: 39, lr: 2.82e-02, grad_scale: 32.0 +2023-03-20 19:03:58,693 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12055.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:04:08,175 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 19:04:10,371 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12076.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:04:13,254 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 19:04:18,836 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 19:04:20,819 INFO [train.py:901] (1/2) Epoch 5, batch 800, loss[loss=0.2271, simple_loss=0.2837, pruned_loss=0.08521, over 7345.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.2907, pruned_loss=0.09206, over 1418472.80 frames. ], batch size: 54, lr: 2.82e-02, grad_scale: 32.0 +2023-03-20 19:04:20,820 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 19:04:24,792 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.858e+02 3.401e+02 4.258e+02 5.414e+02 1.071e+03, threshold=8.517e+02, percent-clipped=1.0 +2023-03-20 19:04:31,264 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 19:04:35,597 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=7.19 vs. limit=5.0 +2023-03-20 19:04:40,847 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12137.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:04:45,700 INFO [train.py:901] (1/2) Epoch 5, batch 850, loss[loss=0.2257, simple_loss=0.2811, pruned_loss=0.08515, over 7272.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.2913, pruned_loss=0.09194, over 1425189.24 frames. ], batch size: 47, lr: 2.81e-02, grad_scale: 32.0 +2023-03-20 19:04:50,323 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 19:04:50,332 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 19:04:55,890 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 19:04:59,484 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12172.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:04:59,944 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 19:05:07,462 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12188.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:05:11,845 INFO [train.py:901] (1/2) Epoch 5, batch 900, loss[loss=0.2178, simple_loss=0.2752, pruned_loss=0.08019, over 7142.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.2911, pruned_loss=0.09218, over 1426158.01 frames. ], batch size: 41, lr: 2.81e-02, grad_scale: 32.0 +2023-03-20 19:05:14,766 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12202.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:05:16,173 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.142e+02 3.418e+02 4.341e+02 5.122e+02 8.381e+02, threshold=8.682e+02, percent-clipped=0.0 +2023-03-20 19:05:20,686 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6033, 4.6451, 4.7251, 5.0059, 5.1482, 5.0488, 4.5553, 4.5114], + device='cuda:1'), covar=tensor([0.0661, 0.1308, 0.1636, 0.1114, 0.0407, 0.1161, 0.0608, 0.0716], + device='cuda:1'), in_proj_covar=tensor([0.0104, 0.0176, 0.0168, 0.0145, 0.0123, 0.0178, 0.0100, 0.0125], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:05:29,669 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6104, 3.7559, 3.5909, 3.8264, 3.1201, 3.7932, 4.0614, 4.1095], + device='cuda:1'), covar=tensor([0.0288, 0.0213, 0.0290, 0.0190, 0.0633, 0.0270, 0.0336, 0.0245], + device='cuda:1'), in_proj_covar=tensor([0.0081, 0.0085, 0.0079, 0.0093, 0.0090, 0.0064, 0.0071, 0.0071], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:05:31,696 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=12236.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:05:33,431 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-20 19:05:37,415 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8109, 2.1583, 2.3142, 2.9201, 1.5614, 2.2330, 1.1120, 3.5034], + device='cuda:1'), covar=tensor([0.0042, 0.0516, 0.1446, 0.0042, 0.3887, 0.0034, 0.0880, 0.0044], + device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0184, 0.0282, 0.0109, 0.0276, 0.0106, 0.0208, 0.0116], + device='cuda:1'), out_proj_covar=tensor([9.9614e-05, 1.6148e-04, 2.2708e-04, 9.8794e-05, 2.3552e-04, 9.7163e-05, + 1.7908e-04, 1.0598e-04], device='cuda:1') +2023-03-20 19:05:37,753 INFO [train.py:901] (1/2) Epoch 5, batch 950, loss[loss=0.2646, simple_loss=0.3147, pruned_loss=0.1072, over 7361.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.2914, pruned_loss=0.09199, over 1430702.52 frames. ], batch size: 65, lr: 2.80e-02, grad_scale: 32.0 +2023-03-20 19:05:38,782 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 19:05:38,882 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1995, 4.4484, 4.2077, 4.3735, 3.8089, 4.5160, 4.7467, 4.7680], + device='cuda:1'), covar=tensor([0.0242, 0.0154, 0.0236, 0.0156, 0.0402, 0.0157, 0.0221, 0.0213], + device='cuda:1'), in_proj_covar=tensor([0.0081, 0.0086, 0.0079, 0.0094, 0.0091, 0.0064, 0.0070, 0.0072], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:05:39,116 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.11 vs. limit=2.0 +2023-03-20 19:05:52,943 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12276.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:05:53,458 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4660, 4.9020, 4.9124, 4.8370, 4.6365, 4.4036, 4.9522, 4.7379], + device='cuda:1'), covar=tensor([0.0417, 0.0330, 0.0522, 0.0502, 0.0345, 0.0356, 0.0361, 0.0534], + device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0088, 0.0096, 0.0080, 0.0072, 0.0095, 0.0084, 0.0075], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:06:00,484 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12291.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:06:01,844 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 19:06:03,331 INFO [train.py:901] (1/2) Epoch 5, batch 1000, loss[loss=0.2553, simple_loss=0.303, pruned_loss=0.1038, over 7274.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.291, pruned_loss=0.09173, over 1432600.24 frames. ], batch size: 70, lr: 2.80e-02, grad_scale: 32.0 +2023-03-20 19:06:06,664 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-03-20 19:06:06,900 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12304.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:06:07,286 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.908e+02 3.426e+02 4.315e+02 5.528e+02 1.209e+03, threshold=8.630e+02, percent-clipped=3.0 +2023-03-20 19:06:11,501 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3227, 3.6911, 3.9619, 3.8795, 3.9081, 3.8939, 4.1921, 3.8882], + device='cuda:1'), covar=tensor([0.0177, 0.0204, 0.0247, 0.0182, 0.0216, 0.0254, 0.0210, 0.0179], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0041, 0.0044, 0.0037, 0.0052, 0.0049, 0.0041, 0.0040], + device='cuda:1'), out_proj_covar=tensor([7.7895e-05, 9.9256e-05, 1.0561e-04, 9.0902e-05, 1.2902e-04, 1.2070e-04, + 1.0485e-04, 9.3144e-05], device='cuda:1') +2023-03-20 19:06:22,529 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 19:06:25,595 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=12339.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:06:26,159 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12340.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:06:29,539 INFO [train.py:901] (1/2) Epoch 5, batch 1050, loss[loss=0.2284, simple_loss=0.2911, pruned_loss=0.08288, over 7282.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.2911, pruned_loss=0.09177, over 1436036.77 frames. ], batch size: 77, lr: 2.79e-02, grad_scale: 32.0 +2023-03-20 19:06:33,684 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12355.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:06:44,517 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 19:06:47,988 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 19:06:50,501 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3710, 4.0845, 3.9710, 3.5464, 3.6639, 2.4123, 1.9516, 4.3455], + device='cuda:1'), covar=tensor([0.0014, 0.0033, 0.0108, 0.0079, 0.0043, 0.0463, 0.0840, 0.0038], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0045, 0.0059, 0.0044, 0.0047, 0.0068, 0.0086, 0.0047], + device='cuda:1'), out_proj_covar=tensor([5.2659e-05, 6.7317e-05, 8.3633e-05, 6.5894e-05, 5.9478e-05, 1.0131e-04, + 1.2902e-04, 6.4310e-05], device='cuda:1') +2023-03-20 19:06:54,357 INFO [train.py:901] (1/2) Epoch 5, batch 1100, loss[loss=0.2278, simple_loss=0.2847, pruned_loss=0.08543, over 7260.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.2903, pruned_loss=0.09126, over 1439718.67 frames. ], batch size: 52, lr: 2.79e-02, grad_scale: 32.0 +2023-03-20 19:06:57,780 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=12403.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:06:58,695 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.093e+02 3.365e+02 4.279e+02 5.255e+02 8.935e+02, threshold=8.557e+02, percent-clipped=2.0 +2023-03-20 19:07:00,396 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6967, 1.6650, 1.7575, 1.9978, 1.9581, 1.7977, 2.3008, 1.9240], + device='cuda:1'), covar=tensor([0.0535, 0.0323, 0.0704, 0.0629, 0.0522, 0.0396, 0.0262, 0.0603], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0032, 0.0032, 0.0032, 0.0033, 0.0036, 0.0031, 0.0032], + device='cuda:1'), out_proj_covar=tensor([7.6780e-05, 6.5463e-05, 6.8319e-05, 6.6296e-05, 6.9985e-05, 7.5232e-05, + 6.8482e-05, 6.6498e-05], device='cuda:1') +2023-03-20 19:07:13,495 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12432.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:07:18,382 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 19:07:18,389 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 19:07:20,469 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12446.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:07:20,807 INFO [train.py:901] (1/2) Epoch 5, batch 1150, loss[loss=0.2467, simple_loss=0.2967, pruned_loss=0.09839, over 7311.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.289, pruned_loss=0.09067, over 1439061.99 frames. ], batch size: 59, lr: 2.78e-02, grad_scale: 32.0 +2023-03-20 19:07:21,080 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 +2023-03-20 19:07:30,247 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 19:07:30,712 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 19:07:33,391 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12472.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:07:35,448 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3027, 1.8156, 1.6927, 2.4612, 1.2954, 2.3892, 1.2256, 2.7840], + device='cuda:1'), covar=tensor([0.0092, 0.0686, 0.2190, 0.0055, 0.4937, 0.0078, 0.1026, 0.0061], + device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0189, 0.0289, 0.0109, 0.0290, 0.0107, 0.0213, 0.0114], + device='cuda:1'), out_proj_covar=tensor([9.9306e-05, 1.6764e-04, 2.3281e-04, 9.8287e-05, 2.4674e-04, 9.9585e-05, + 1.8379e-04, 1.0218e-04], device='cuda:1') +2023-03-20 19:07:45,729 INFO [train.py:901] (1/2) Epoch 5, batch 1200, loss[loss=0.251, simple_loss=0.3033, pruned_loss=0.09935, over 7309.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.2888, pruned_loss=0.09067, over 1439192.51 frames. ], batch size: 80, lr: 2.78e-02, grad_scale: 32.0 +2023-03-20 19:07:48,926 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12502.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:07:50,272 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.922e+02 3.243e+02 4.028e+02 5.026e+02 9.782e+02, threshold=8.056e+02, percent-clipped=2.0 +2023-03-20 19:07:51,445 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12507.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:07:58,504 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=12520.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:08:05,443 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 19:08:11,906 INFO [train.py:901] (1/2) Epoch 5, batch 1250, loss[loss=0.2441, simple_loss=0.2995, pruned_loss=0.09433, over 7306.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.2883, pruned_loss=0.09056, over 1439542.66 frames. ], batch size: 83, lr: 2.77e-02, grad_scale: 32.0 +2023-03-20 19:08:13,464 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=12550.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:08:26,598 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12576.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:08:28,033 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 19:08:32,658 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 19:08:33,664 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 19:08:37,623 INFO [train.py:901] (1/2) Epoch 5, batch 1300, loss[loss=0.2251, simple_loss=0.2837, pruned_loss=0.08324, over 7162.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.2887, pruned_loss=0.09078, over 1439286.92 frames. ], batch size: 41, lr: 2.77e-02, grad_scale: 32.0 +2023-03-20 19:08:38,471 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 +2023-03-20 19:08:42,148 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12604.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:08:42,499 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.705e+02 3.330e+02 4.266e+02 5.499e+02 1.083e+03, threshold=8.532e+02, percent-clipped=4.0 +2023-03-20 19:08:50,443 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-03-20 19:08:52,194 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=12624.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:08:54,277 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6862, 3.4130, 3.3315, 3.3961, 3.1795, 3.2886, 3.5494, 3.1047], + device='cuda:1'), covar=tensor([0.0191, 0.0200, 0.0232, 0.0227, 0.0332, 0.0273, 0.0349, 0.0343], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0036, 0.0039, 0.0033, 0.0048, 0.0043, 0.0035, 0.0036], + device='cuda:1'), out_proj_covar=tensor([7.2600e-05, 8.6746e-05, 9.0961e-05, 8.1662e-05, 1.1839e-04, 1.0629e-04, + 9.1619e-05, 8.5326e-05], device='cuda:1') +2023-03-20 19:08:58,121 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 19:09:00,172 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12640.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:09:00,555 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 19:09:03,583 INFO [train.py:901] (1/2) Epoch 5, batch 1350, loss[loss=0.26, simple_loss=0.3122, pruned_loss=0.1039, over 7278.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.2893, pruned_loss=0.09115, over 1438357.78 frames. ], batch size: 66, lr: 2.77e-02, grad_scale: 32.0 +2023-03-20 19:09:04,099 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 19:09:06,149 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=12652.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:09:13,540 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 19:09:24,676 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=12688.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:09:29,644 INFO [train.py:901] (1/2) Epoch 5, batch 1400, loss[loss=0.245, simple_loss=0.2948, pruned_loss=0.09762, over 7270.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.2894, pruned_loss=0.09096, over 1439606.15 frames. ], batch size: 52, lr: 2.76e-02, grad_scale: 32.0 +2023-03-20 19:09:30,263 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8682, 3.6182, 3.4377, 3.0660, 3.2146, 2.2209, 1.6345, 3.7889], + device='cuda:1'), covar=tensor([0.0018, 0.0074, 0.0103, 0.0108, 0.0058, 0.0489, 0.0841, 0.0048], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0047, 0.0060, 0.0048, 0.0050, 0.0071, 0.0088, 0.0050], + device='cuda:1'), out_proj_covar=tensor([5.4578e-05, 6.9512e-05, 8.5210e-05, 7.0721e-05, 6.4462e-05, 1.0574e-04, + 1.3196e-04, 6.9155e-05], device='cuda:1') +2023-03-20 19:09:33,686 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.299e+02 3.528e+02 4.338e+02 5.415e+02 8.083e+02, threshold=8.676e+02, percent-clipped=0.0 +2023-03-20 19:09:33,850 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7573, 1.6251, 1.3449, 1.8083, 1.9410, 1.6578, 2.0851, 1.6387], + device='cuda:1'), covar=tensor([0.0707, 0.0867, 0.0715, 0.0470, 0.0556, 0.0611, 0.0374, 0.0594], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0034, 0.0034, 0.0032, 0.0034, 0.0037, 0.0031, 0.0034], + device='cuda:1'), out_proj_covar=tensor([7.9835e-05, 6.9265e-05, 7.1449e-05, 6.5750e-05, 7.2941e-05, 7.7738e-05, + 6.9847e-05, 7.1270e-05], device='cuda:1') +2023-03-20 19:09:40,833 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2756, 4.4046, 4.0677, 4.2131, 3.9401, 4.4055, 4.5302, 4.7318], + device='cuda:1'), covar=tensor([0.0173, 0.0143, 0.0185, 0.0204, 0.0327, 0.0161, 0.0252, 0.0150], + device='cuda:1'), in_proj_covar=tensor([0.0081, 0.0084, 0.0078, 0.0094, 0.0086, 0.0066, 0.0067, 0.0069], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:09:47,220 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 19:09:47,297 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12732.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:09:54,830 INFO [train.py:901] (1/2) Epoch 5, batch 1450, loss[loss=0.225, simple_loss=0.2913, pruned_loss=0.07938, over 7303.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.2894, pruned_loss=0.0905, over 1441138.94 frames. ], batch size: 68, lr: 2.76e-02, grad_scale: 32.0 +2023-03-20 19:10:01,051 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4048, 3.3665, 3.4211, 3.4331, 3.1165, 3.4165, 3.6818, 3.8097], + device='cuda:1'), covar=tensor([0.0323, 0.0271, 0.0288, 0.0274, 0.0472, 0.0409, 0.0336, 0.0203], + device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0085, 0.0078, 0.0093, 0.0086, 0.0066, 0.0067, 0.0070], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:10:03,119 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12763.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:10:05,962 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 +2023-03-20 19:10:11,860 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 19:10:12,890 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=12780.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:10:21,344 INFO [train.py:901] (1/2) Epoch 5, batch 1500, loss[loss=0.2495, simple_loss=0.2982, pruned_loss=0.1004, over 7283.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.29, pruned_loss=0.0904, over 1440715.10 frames. ], batch size: 66, lr: 2.75e-02, grad_scale: 32.0 +2023-03-20 19:10:24,159 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12802.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:10:25,595 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 3.496e+02 4.325e+02 5.786e+02 9.440e+02, threshold=8.649e+02, percent-clipped=4.0 +2023-03-20 19:10:28,155 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 19:10:35,406 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12824.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 19:10:46,767 INFO [train.py:901] (1/2) Epoch 5, batch 1550, loss[loss=0.2365, simple_loss=0.2965, pruned_loss=0.08828, over 7360.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.2892, pruned_loss=0.0905, over 1440110.02 frames. ], batch size: 51, lr: 2.75e-02, grad_scale: 32.0 +2023-03-20 19:10:52,221 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 19:11:13,173 INFO [train.py:901] (1/2) Epoch 5, batch 1600, loss[loss=0.2545, simple_loss=0.3096, pruned_loss=0.0997, over 7296.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.2882, pruned_loss=0.0896, over 1439603.75 frames. ], batch size: 86, lr: 2.74e-02, grad_scale: 32.0 +2023-03-20 19:11:17,086 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.145e+02 3.367e+02 4.098e+02 5.521e+02 1.235e+03, threshold=8.196e+02, percent-clipped=2.0 +2023-03-20 19:11:20,328 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3350, 2.2296, 2.6117, 3.3322, 2.7979, 3.3489, 2.7846, 3.2862], + device='cuda:1'), covar=tensor([0.0786, 0.0417, 0.1011, 0.0092, 0.0027, 0.0029, 0.0021, 0.0078], + device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0174, 0.0221, 0.0131, 0.0090, 0.0089, 0.0085, 0.0092], + device='cuda:1'), out_proj_covar=tensor([2.1250e-04, 1.6523e-04, 1.9980e-04, 1.2861e-04, 8.7597e-05, 8.3153e-05, + 8.4422e-05, 9.2412e-05], device='cuda:1') +2023-03-20 19:11:22,216 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 19:11:23,217 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 19:11:26,768 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 19:11:36,606 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 19:11:38,693 INFO [train.py:901] (1/2) Epoch 5, batch 1650, loss[loss=0.2419, simple_loss=0.2959, pruned_loss=0.09399, over 7361.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.2887, pruned_loss=0.08997, over 1439815.48 frames. ], batch size: 73, lr: 2.74e-02, grad_scale: 32.0 +2023-03-20 19:11:40,905 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1225, 1.4347, 0.7073, 1.0415, 1.1019, 1.3198, 1.2701, 0.9854], + device='cuda:1'), covar=tensor([0.0363, 0.0463, 0.0356, 0.0237, 0.0370, 0.0374, 0.0260, 0.0422], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0021, 0.0019, 0.0019, 0.0020, 0.0019, 0.0019, 0.0019], + device='cuda:1'), out_proj_covar=tensor([4.0885e-05, 4.3772e-05, 3.5147e-05, 3.4591e-05, 4.1403e-05, 3.9898e-05, + 3.7613e-05, 4.1296e-05], device='cuda:1') +2023-03-20 19:11:41,282 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 19:11:49,451 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 19:12:04,412 INFO [train.py:901] (1/2) Epoch 5, batch 1700, loss[loss=0.2089, simple_loss=0.2631, pruned_loss=0.07736, over 7347.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.2886, pruned_loss=0.08982, over 1439429.63 frames. ], batch size: 54, lr: 2.73e-02, grad_scale: 32.0 +2023-03-20 19:12:07,245 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 19:12:08,703 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.127e+02 3.688e+02 4.669e+02 5.560e+02 1.145e+03, threshold=9.338e+02, percent-clipped=6.0 +2023-03-20 19:12:10,782 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 19:12:20,811 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 19:12:21,921 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3962, 3.2620, 2.7464, 2.3463, 2.9536, 3.0110, 2.0413, 2.9336], + device='cuda:1'), covar=tensor([0.1259, 0.0124, 0.1532, 0.2257, 0.0570, 0.0847, 0.1749, 0.0774], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0022, 0.0025, 0.0026, 0.0021, 0.0022, 0.0030, 0.0024], + device='cuda:1'), out_proj_covar=tensor([5.7554e-05, 4.6585e-05, 6.0776e-05, 6.0959e-05, 5.2896e-05, 5.5442e-05, + 6.9830e-05, 5.4411e-05], device='cuda:1') +2023-03-20 19:12:31,115 INFO [train.py:901] (1/2) Epoch 5, batch 1750, loss[loss=0.226, simple_loss=0.2714, pruned_loss=0.09034, over 7244.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.2897, pruned_loss=0.09098, over 1437332.82 frames. ], batch size: 45, lr: 2.73e-02, grad_scale: 32.0 +2023-03-20 19:12:38,185 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-03-20 19:12:45,668 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 19:12:46,677 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 19:12:56,138 INFO [train.py:901] (1/2) Epoch 5, batch 1800, loss[loss=0.2204, simple_loss=0.2823, pruned_loss=0.07921, over 7345.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.2885, pruned_loss=0.09022, over 1439674.77 frames. ], batch size: 63, lr: 2.72e-02, grad_scale: 32.0 +2023-03-20 19:12:58,723 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13102.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:13:00,051 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 +2023-03-20 19:13:00,109 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.985e+02 3.231e+02 4.030e+02 4.960e+02 9.791e+02, threshold=8.060e+02, percent-clipped=2.0 +2023-03-20 19:13:07,148 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=13119.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 19:13:08,161 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 19:13:21,151 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 19:13:21,608 INFO [train.py:901] (1/2) Epoch 5, batch 1850, loss[loss=0.259, simple_loss=0.3139, pruned_loss=0.1021, over 7296.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.2898, pruned_loss=0.09078, over 1439664.23 frames. ], batch size: 68, lr: 2.72e-02, grad_scale: 32.0 +2023-03-20 19:13:23,148 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=13150.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:13:31,163 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 19:13:46,564 INFO [train.py:901] (1/2) Epoch 5, batch 1900, loss[loss=0.2649, simple_loss=0.3147, pruned_loss=0.1075, over 7326.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.2896, pruned_loss=0.09069, over 1441464.79 frames. ], batch size: 61, lr: 2.71e-02, grad_scale: 32.0 +2023-03-20 19:13:48,137 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 19:13:50,870 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.920e+02 3.572e+02 4.381e+02 5.635e+02 1.189e+03, threshold=8.761e+02, percent-clipped=5.0 +2023-03-20 19:14:00,916 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8272, 3.5336, 3.4948, 3.3473, 3.4407, 3.5296, 3.7562, 3.3942], + device='cuda:1'), covar=tensor([0.0092, 0.0134, 0.0130, 0.0174, 0.0180, 0.0111, 0.0149, 0.0136], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0035, 0.0041, 0.0033, 0.0050, 0.0045, 0.0040, 0.0038], + device='cuda:1'), out_proj_covar=tensor([7.9681e-05, 8.9335e-05, 9.8089e-05, 8.2227e-05, 1.2740e-04, 1.1383e-04, + 1.0517e-04, 9.0513e-05], device='cuda:1') +2023-03-20 19:14:13,360 INFO [train.py:901] (1/2) Epoch 5, batch 1950, loss[loss=0.2289, simple_loss=0.2818, pruned_loss=0.08803, over 7281.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.2888, pruned_loss=0.09052, over 1440919.52 frames. ], batch size: 77, lr: 2.71e-02, grad_scale: 32.0 +2023-03-20 19:14:13,866 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 19:14:24,445 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 19:14:29,415 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 19:14:29,953 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 19:14:38,264 INFO [train.py:901] (1/2) Epoch 5, batch 2000, loss[loss=0.3259, simple_loss=0.3634, pruned_loss=0.1441, over 6765.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.2884, pruned_loss=0.08991, over 1443027.66 frames. ], batch size: 107, lr: 2.71e-02, grad_scale: 32.0 +2023-03-20 19:14:43,472 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.925e+02 3.298e+02 4.123e+02 5.170e+02 8.355e+02, threshold=8.246e+02, percent-clipped=0.0 +2023-03-20 19:14:47,055 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 19:14:51,383 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.09 vs. limit=2.0 +2023-03-20 19:14:57,649 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 19:15:04,689 INFO [train.py:901] (1/2) Epoch 5, batch 2050, loss[loss=0.2166, simple_loss=0.2794, pruned_loss=0.07685, over 7319.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.2886, pruned_loss=0.0901, over 1443440.96 frames. ], batch size: 83, lr: 2.70e-02, grad_scale: 32.0 +2023-03-20 19:15:05,208 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 19:15:15,850 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7719, 2.2975, 2.1187, 3.1446, 1.5033, 2.6897, 1.2791, 2.8791], + device='cuda:1'), covar=tensor([0.0048, 0.0524, 0.1869, 0.0040, 0.4356, 0.0071, 0.0897, 0.0055], + device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0197, 0.0287, 0.0112, 0.0293, 0.0117, 0.0223, 0.0125], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 19:15:30,842 INFO [train.py:901] (1/2) Epoch 5, batch 2100, loss[loss=0.2365, simple_loss=0.2921, pruned_loss=0.0904, over 7293.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.2889, pruned_loss=0.09024, over 1441913.51 frames. ], batch size: 68, lr: 2.70e-02, grad_scale: 32.0 +2023-03-20 19:15:32,831 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6780, 3.4226, 3.3170, 3.5824, 2.7024, 2.2331, 3.5483, 2.8535], + device='cuda:1'), covar=tensor([0.0070, 0.0058, 0.0072, 0.0023, 0.0139, 0.0268, 0.0082, 0.0284], + device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0140, 0.0195, 0.0120, 0.0240, 0.0261, 0.0170, 0.0270], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 19:15:35,086 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.118e+02 3.365e+02 4.362e+02 5.459e+02 9.984e+02, threshold=8.723e+02, percent-clipped=1.0 +2023-03-20 19:15:39,095 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 19:15:42,126 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 19:15:42,236 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13419.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:15:56,033 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-03-20 19:15:56,128 INFO [train.py:901] (1/2) Epoch 5, batch 2150, loss[loss=0.2446, simple_loss=0.2909, pruned_loss=0.09918, over 7346.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.2889, pruned_loss=0.08969, over 1443367.99 frames. ], batch size: 54, lr: 2.69e-02, grad_scale: 32.0 +2023-03-20 19:15:58,749 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.1322, 5.4626, 5.4629, 5.3827, 5.0567, 4.9732, 5.5471, 5.3985], + device='cuda:1'), covar=tensor([0.0261, 0.0230, 0.0423, 0.0431, 0.0322, 0.0210, 0.0241, 0.0320], + device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0096, 0.0101, 0.0088, 0.0080, 0.0102, 0.0091, 0.0078], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:16:00,231 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.1377, 5.5202, 5.5613, 5.3419, 5.0852, 5.0314, 5.5526, 5.4010], + device='cuda:1'), covar=tensor([0.0234, 0.0191, 0.0273, 0.0425, 0.0304, 0.0203, 0.0217, 0.0303], + device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0096, 0.0101, 0.0088, 0.0080, 0.0102, 0.0091, 0.0078], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:16:06,800 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=13467.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:16:22,386 INFO [train.py:901] (1/2) Epoch 5, batch 2200, loss[loss=0.1928, simple_loss=0.2484, pruned_loss=0.0686, over 7156.00 frames. ], tot_loss[loss=0.234, simple_loss=0.2887, pruned_loss=0.08964, over 1444831.43 frames. ], batch size: 41, lr: 2.69e-02, grad_scale: 32.0 +2023-03-20 19:16:26,404 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 3.587e+02 4.467e+02 5.823e+02 1.129e+03, threshold=8.934e+02, percent-clipped=3.0 +2023-03-20 19:16:26,465 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 19:16:47,779 INFO [train.py:901] (1/2) Epoch 5, batch 2250, loss[loss=0.2322, simple_loss=0.2913, pruned_loss=0.08655, over 7316.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.2897, pruned_loss=0.08996, over 1445258.34 frames. ], batch size: 83, lr: 2.68e-02, grad_scale: 32.0 +2023-03-20 19:16:50,556 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1470, 0.8702, 1.4229, 1.1533, 1.2065, 0.6252, 0.4915, 0.8651], + device='cuda:1'), covar=tensor([0.0521, 0.0928, 0.0324, 0.0137, 0.0632, 0.1145, 0.0378, 0.0542], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0035, 0.0027, 0.0027, 0.0027, 0.0028, 0.0030, 0.0029], + device='cuda:1'), out_proj_covar=tensor([4.5024e-05, 6.4367e-05, 4.0017e-05, 4.1465e-05, 4.4156e-05, 4.9314e-05, + 4.8139e-05, 4.8596e-05], device='cuda:1') +2023-03-20 19:17:01,116 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 19:17:01,589 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 19:17:08,027 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.90 vs. limit=2.0 +2023-03-20 19:17:13,139 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 19:17:14,170 INFO [train.py:901] (1/2) Epoch 5, batch 2300, loss[loss=0.2313, simple_loss=0.2848, pruned_loss=0.08884, over 7321.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.289, pruned_loss=0.08957, over 1442812.36 frames. ], batch size: 59, lr: 2.68e-02, grad_scale: 32.0 +2023-03-20 19:17:18,480 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.119e+02 3.634e+02 4.317e+02 5.619e+02 1.145e+03, threshold=8.633e+02, percent-clipped=2.0 +2023-03-20 19:17:26,733 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9490, 2.9573, 2.2215, 2.3951, 2.5337, 2.7909, 1.8694, 2.9693], + device='cuda:1'), covar=tensor([0.0510, 0.0270, 0.2230, 0.1937, 0.1019, 0.1295, 0.2653, 0.0774], + device='cuda:1'), in_proj_covar=tensor([0.0023, 0.0024, 0.0026, 0.0027, 0.0024, 0.0026, 0.0030, 0.0025], + device='cuda:1'), out_proj_covar=tensor([5.7780e-05, 5.3825e-05, 6.4113e-05, 6.4891e-05, 5.9528e-05, 6.4729e-05, + 7.2663e-05, 6.0839e-05], device='cuda:1') +2023-03-20 19:17:40,826 INFO [train.py:901] (1/2) Epoch 5, batch 2350, loss[loss=0.2302, simple_loss=0.2894, pruned_loss=0.08548, over 7252.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.2889, pruned_loss=0.08977, over 1444347.38 frames. ], batch size: 55, lr: 2.68e-02, grad_scale: 32.0 +2023-03-20 19:18:00,343 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 19:18:05,789 INFO [train.py:901] (1/2) Epoch 5, batch 2400, loss[loss=0.2624, simple_loss=0.3162, pruned_loss=0.1042, over 6711.00 frames. ], tot_loss[loss=0.233, simple_loss=0.2878, pruned_loss=0.08913, over 1442489.62 frames. ], batch size: 107, lr: 2.67e-02, grad_scale: 32.0 +2023-03-20 19:18:05,827 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 19:18:09,735 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.037e+02 3.360e+02 4.208e+02 5.531e+02 1.498e+03, threshold=8.415e+02, percent-clipped=3.0 +2023-03-20 19:18:16,861 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 19:18:19,435 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 19:18:27,304 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=13737.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:18:32,094 INFO [train.py:901] (1/2) Epoch 5, batch 2450, loss[loss=0.2361, simple_loss=0.295, pruned_loss=0.08862, over 7330.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.2889, pruned_loss=0.08979, over 1442510.09 frames. ], batch size: 61, lr: 2.67e-02, grad_scale: 32.0 +2023-03-20 19:18:45,566 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 19:18:57,133 INFO [train.py:901] (1/2) Epoch 5, batch 2500, loss[loss=0.2112, simple_loss=0.2673, pruned_loss=0.07757, over 7201.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.2873, pruned_loss=0.08899, over 1441432.94 frames. ], batch size: 39, lr: 2.66e-02, grad_scale: 32.0 +2023-03-20 19:18:57,790 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=13798.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 19:19:01,980 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.868e+02 3.607e+02 4.140e+02 5.298e+02 1.184e+03, threshold=8.279e+02, percent-clipped=3.0 +2023-03-20 19:19:12,224 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 19:19:21,423 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5886, 3.7190, 2.4589, 3.8777, 3.4580, 3.5748, 2.7095, 2.4342], + device='cuda:1'), covar=tensor([0.0038, 0.0149, 0.0442, 0.0079, 0.0046, 0.0071, 0.0721, 0.0443], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0129, 0.0231, 0.0118, 0.0118, 0.0117, 0.0236, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0001, 0.0001, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 19:19:23,800 INFO [train.py:901] (1/2) Epoch 5, batch 2550, loss[loss=0.1737, simple_loss=0.2378, pruned_loss=0.05477, over 7156.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.2869, pruned_loss=0.08863, over 1440820.99 frames. ], batch size: 39, lr: 2.66e-02, grad_scale: 32.0 +2023-03-20 19:19:28,592 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-03-20 19:19:48,486 INFO [train.py:901] (1/2) Epoch 5, batch 2600, loss[loss=0.2364, simple_loss=0.2934, pruned_loss=0.08968, over 7265.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.287, pruned_loss=0.08896, over 1439570.06 frames. ], batch size: 64, lr: 2.65e-02, grad_scale: 16.0 +2023-03-20 19:19:52,792 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.171e+02 3.332e+02 4.151e+02 5.370e+02 8.720e+02, threshold=8.301e+02, percent-clipped=2.0 +2023-03-20 19:20:13,543 INFO [train.py:901] (1/2) Epoch 5, batch 2650, loss[loss=0.2136, simple_loss=0.2517, pruned_loss=0.08773, over 6941.00 frames. ], tot_loss[loss=0.232, simple_loss=0.2866, pruned_loss=0.08867, over 1440178.95 frames. ], batch size: 35, lr: 2.65e-02, grad_scale: 16.0 +2023-03-20 19:20:21,148 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4398, 1.8551, 1.4688, 1.7787, 1.7966, 1.5844, 1.8405, 1.5712], + device='cuda:1'), covar=tensor([0.1043, 0.0638, 0.1298, 0.0607, 0.0879, 0.0710, 0.0849, 0.0805], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0033, 0.0035, 0.0034, 0.0036, 0.0036, 0.0032, 0.0032], + device='cuda:1'), out_proj_covar=tensor([8.8266e-05, 7.4380e-05, 7.8507e-05, 7.6381e-05, 8.1512e-05, 8.0703e-05, + 7.6157e-05, 7.3669e-05], device='cuda:1') +2023-03-20 19:20:30,688 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.07 vs. limit=5.0 +2023-03-20 19:20:33,349 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2547, 3.2197, 2.3360, 3.4948, 2.8885, 3.3429, 2.3237, 2.0087], + device='cuda:1'), covar=tensor([0.0020, 0.0369, 0.0503, 0.0118, 0.0076, 0.0075, 0.0716, 0.0517], + device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0130, 0.0239, 0.0120, 0.0120, 0.0118, 0.0236, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0001, 0.0001, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 19:20:38,634 INFO [train.py:901] (1/2) Epoch 5, batch 2700, loss[loss=0.2119, simple_loss=0.2792, pruned_loss=0.07226, over 7350.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.2864, pruned_loss=0.08845, over 1441143.21 frames. ], batch size: 51, lr: 2.65e-02, grad_scale: 16.0 +2023-03-20 19:20:43,427 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.392e+02 3.598e+02 4.306e+02 4.916e+02 1.230e+03, threshold=8.611e+02, percent-clipped=5.0 +2023-03-20 19:21:03,777 INFO [train.py:901] (1/2) Epoch 5, batch 2750, loss[loss=0.2503, simple_loss=0.3037, pruned_loss=0.09847, over 7340.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.2859, pruned_loss=0.08781, over 1438975.50 frames. ], batch size: 83, lr: 2.64e-02, grad_scale: 16.0 +2023-03-20 19:21:16,266 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14072.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:21:26,577 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14093.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 19:21:28,448 INFO [train.py:901] (1/2) Epoch 5, batch 2800, loss[loss=0.2371, simple_loss=0.2913, pruned_loss=0.09145, over 7272.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.2854, pruned_loss=0.08767, over 1436467.41 frames. ], batch size: 64, lr: 2.64e-02, grad_scale: 16.0 +2023-03-20 19:21:31,501 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14103.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:21:32,842 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.456e+02 3.620e+02 4.147e+02 5.352e+02 9.706e+02, threshold=8.293e+02, percent-clipped=5.0 +2023-03-20 19:21:53,269 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 19:21:53,321 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 19:21:53,345 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 19:21:53,361 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 19:21:53,836 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 19:21:53,856 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 19:21:53,868 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 19:21:53,964 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 19:21:54,038 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 19:21:54,052 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 19:21:54,069 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 19:21:54,235 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 19:21:54,242 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 19:21:54,249 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 19:21:54,390 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 19:21:54,393 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 19:21:54,451 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 19:22:01,134 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14120.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 19:22:01,488 INFO [train.py:901] (1/2) Epoch 6, batch 0, loss[loss=0.2167, simple_loss=0.2775, pruned_loss=0.07793, over 7329.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2775, pruned_loss=0.07793, over 7329.00 frames. ], batch size: 75, lr: 2.53e-02, grad_scale: 16.0 +2023-03-20 19:22:01,488 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 19:22:17,875 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0177, 1.2144, 0.8477, 0.8846, 1.0177, 0.8807, 0.8222, 1.0572], + device='cuda:1'), covar=tensor([0.0186, 0.0315, 0.0460, 0.0116, 0.0340, 0.0236, 0.0515, 0.0109], + device='cuda:1'), in_proj_covar=tensor([0.0017, 0.0019, 0.0016, 0.0016, 0.0017, 0.0017, 0.0016, 0.0017], + device='cuda:1'), out_proj_covar=tensor([2.0441e-05, 2.1802e-05, 2.2012e-05, 1.9686e-05, 1.9756e-05, 2.2285e-05, + 2.1194e-05, 2.4836e-05], device='cuda:1') +2023-03-20 19:22:21,099 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3590, 3.5055, 3.2228, 3.1011, 3.0069, 2.1578, 1.4857, 3.6190], + device='cuda:1'), covar=tensor([0.0040, 0.0038, 0.0061, 0.0075, 0.0048, 0.0576, 0.1028, 0.0055], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0047, 0.0066, 0.0052, 0.0054, 0.0081, 0.0096, 0.0053], + device='cuda:1'), out_proj_covar=tensor([6.2659e-05, 7.3514e-05, 9.5292e-05, 7.7542e-05, 7.1343e-05, 1.2097e-04, + 1.4199e-04, 7.3696e-05], device='cuda:1') +2023-03-20 19:22:27,062 INFO [train.py:935] (1/2) Epoch 6, validation: loss=0.1912, simple_loss=0.2765, pruned_loss=0.05298, over 1622729.00 frames. +2023-03-20 19:22:27,063 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-20 19:22:33,270 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14133.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:22:34,131 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 19:22:44,803 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 19:22:49,678 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.25 vs. limit=5.0 +2023-03-20 19:22:50,049 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14164.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:22:51,909 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 19:22:53,348 INFO [train.py:901] (1/2) Epoch 6, batch 50, loss[loss=0.2083, simple_loss=0.2669, pruned_loss=0.07489, over 7264.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.2882, pruned_loss=0.08765, over 325993.50 frames. ], batch size: 47, lr: 2.53e-02, grad_scale: 16.0 +2023-03-20 19:22:53,873 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 19:22:56,406 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 19:22:58,609 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14181.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 19:23:11,287 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.057e+02 3.549e+02 4.505e+02 5.782e+02 1.054e+03, threshold=9.011e+02, percent-clipped=3.0 +2023-03-20 19:23:11,482 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.9693, 1.1485, 0.8007, 0.8224, 0.9196, 0.7882, 0.6869, 0.9917], + device='cuda:1'), covar=tensor([0.0237, 0.0121, 0.0295, 0.0108, 0.0108, 0.0216, 0.0207, 0.0172], + device='cuda:1'), in_proj_covar=tensor([0.0016, 0.0017, 0.0014, 0.0015, 0.0016, 0.0016, 0.0016, 0.0016], + device='cuda:1'), out_proj_covar=tensor([2.0477e-05, 2.0579e-05, 2.0370e-05, 1.8603e-05, 1.8521e-05, 2.1240e-05, + 2.0729e-05, 2.3522e-05], device='cuda:1') +2023-03-20 19:23:13,334 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 19:23:13,813 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 19:23:16,383 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.8247, 0.6985, 1.1314, 0.7181, 0.8114, 0.7717, 0.8686, 0.9544], + device='cuda:1'), covar=tensor([0.0371, 0.0852, 0.0228, 0.0248, 0.0655, 0.0525, 0.0298, 0.0573], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0035, 0.0028, 0.0028, 0.0027, 0.0026, 0.0032, 0.0029], + device='cuda:1'), out_proj_covar=tensor([4.8345e-05, 6.6264e-05, 4.0153e-05, 4.2832e-05, 4.6467e-05, 4.7963e-05, + 5.0679e-05, 5.0636e-05], device='cuda:1') +2023-03-20 19:23:18,728 INFO [train.py:901] (1/2) Epoch 6, batch 100, loss[loss=0.2572, simple_loss=0.3059, pruned_loss=0.1043, over 7317.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.2884, pruned_loss=0.08795, over 575848.90 frames. ], batch size: 80, lr: 2.52e-02, grad_scale: 16.0 +2023-03-20 19:23:30,300 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-03-20 19:23:38,781 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.39 vs. limit=5.0 +2023-03-20 19:23:45,033 INFO [train.py:901] (1/2) Epoch 6, batch 150, loss[loss=0.2319, simple_loss=0.2874, pruned_loss=0.0882, over 7316.00 frames. ], tot_loss[loss=0.228, simple_loss=0.2841, pruned_loss=0.08595, over 767870.74 frames. ], batch size: 75, lr: 2.52e-02, grad_scale: 16.0 +2023-03-20 19:24:02,563 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.799e+02 3.314e+02 3.960e+02 4.953e+02 9.074e+02, threshold=7.919e+02, percent-clipped=1.0 +2023-03-20 19:24:10,049 INFO [train.py:901] (1/2) Epoch 6, batch 200, loss[loss=0.257, simple_loss=0.3034, pruned_loss=0.1053, over 7343.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.284, pruned_loss=0.08557, over 918310.68 frames. ], batch size: 73, lr: 2.52e-02, grad_scale: 16.0 +2023-03-20 19:24:16,760 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 19:24:21,795 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 19:24:27,227 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 19:24:36,317 INFO [train.py:901] (1/2) Epoch 6, batch 250, loss[loss=0.223, simple_loss=0.2855, pruned_loss=0.08025, over 7309.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.2837, pruned_loss=0.08508, over 1034746.37 frames. ], batch size: 49, lr: 2.51e-02, grad_scale: 16.0 +2023-03-20 19:24:39,893 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0245, 3.5558, 3.8214, 3.5659, 3.7104, 3.7019, 3.8754, 3.7776], + device='cuda:1'), covar=tensor([0.0100, 0.0145, 0.0130, 0.0167, 0.0173, 0.0127, 0.0157, 0.0108], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0041, 0.0046, 0.0039, 0.0056, 0.0050, 0.0044, 0.0040], + device='cuda:1'), out_proj_covar=tensor([8.5345e-05, 1.0797e-04, 1.1483e-04, 9.5392e-05, 1.4862e-04, 1.3006e-04, + 1.1989e-04, 9.5600e-05], device='cuda:1') +2023-03-20 19:24:40,324 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 19:24:43,143 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 +2023-03-20 19:24:46,377 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1905, 4.0227, 3.6959, 3.3079, 3.6588, 2.4899, 1.8961, 4.0342], + device='cuda:1'), covar=tensor([0.0015, 0.0019, 0.0071, 0.0072, 0.0020, 0.0379, 0.0701, 0.0034], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0045, 0.0066, 0.0053, 0.0052, 0.0079, 0.0095, 0.0053], + device='cuda:1'), out_proj_covar=tensor([5.9435e-05, 7.0597e-05, 9.5590e-05, 7.8978e-05, 6.9196e-05, 1.1786e-04, + 1.4005e-04, 7.4171e-05], device='cuda:1') +2023-03-20 19:24:47,351 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14393.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:24:54,059 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 3.223e+02 4.250e+02 5.508e+02 1.036e+03, threshold=8.501e+02, percent-clipped=7.0 +2023-03-20 19:24:58,265 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14413.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:25:01,758 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 19:25:02,709 INFO [train.py:901] (1/2) Epoch 6, batch 300, loss[loss=0.2237, simple_loss=0.2892, pruned_loss=0.07915, over 7321.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.2848, pruned_loss=0.08578, over 1125164.41 frames. ], batch size: 59, lr: 2.51e-02, grad_scale: 16.0 +2023-03-20 19:25:06,329 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14428.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:25:10,229 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 19:25:12,839 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=14441.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:25:13,338 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1599, 4.4113, 4.5157, 4.4225, 4.3977, 4.0005, 4.4554, 4.3923], + device='cuda:1'), covar=tensor([0.0310, 0.0367, 0.0364, 0.0456, 0.0376, 0.0324, 0.0365, 0.0491], + device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0103, 0.0097, 0.0086, 0.0079, 0.0106, 0.0094, 0.0079], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:25:16,076 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.12 vs. limit=5.0 +2023-03-20 19:25:21,778 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14459.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:25:27,659 INFO [train.py:901] (1/2) Epoch 6, batch 350, loss[loss=0.181, simple_loss=0.2196, pruned_loss=0.07117, over 6553.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.2838, pruned_loss=0.08484, over 1196225.43 frames. ], batch size: 28, lr: 2.50e-02, grad_scale: 16.0 +2023-03-20 19:25:29,269 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14474.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:25:30,186 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14476.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 19:25:35,724 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14487.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:25:43,733 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 19:25:46,321 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.128e+02 3.102e+02 3.888e+02 4.761e+02 1.038e+03, threshold=7.777e+02, percent-clipped=1.0 +2023-03-20 19:25:48,341 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8470, 2.0767, 2.0563, 1.9125, 1.8760, 1.9275, 2.2897, 2.0518], + device='cuda:1'), covar=tensor([0.0729, 0.0452, 0.0618, 0.0693, 0.0875, 0.0391, 0.0688, 0.0687], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0033, 0.0034, 0.0035, 0.0034, 0.0035, 0.0032, 0.0033], + device='cuda:1'), out_proj_covar=tensor([8.6174e-05, 7.5754e-05, 7.8841e-05, 7.8249e-05, 8.0399e-05, 7.8903e-05, + 7.6417e-05, 7.3846e-05], device='cuda:1') +2023-03-20 19:25:53,705 INFO [train.py:901] (1/2) Epoch 6, batch 400, loss[loss=0.2107, simple_loss=0.2483, pruned_loss=0.08652, over 7046.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.2834, pruned_loss=0.08462, over 1250414.17 frames. ], batch size: 35, lr: 2.50e-02, grad_scale: 16.0 +2023-03-20 19:26:07,235 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14548.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:26:09,696 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5511, 4.8808, 4.8871, 4.8104, 4.6266, 4.3403, 4.9001, 4.7304], + device='cuda:1'), covar=tensor([0.0294, 0.0381, 0.0513, 0.0467, 0.0289, 0.0307, 0.0363, 0.0426], + device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0104, 0.0100, 0.0087, 0.0078, 0.0106, 0.0091, 0.0081], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:26:18,016 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4470, 4.5384, 4.5251, 4.8745, 5.0230, 4.9110, 4.3042, 4.3392], + device='cuda:1'), covar=tensor([0.0797, 0.1439, 0.1501, 0.0758, 0.0361, 0.0975, 0.0505, 0.0810], + device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0191, 0.0180, 0.0157, 0.0128, 0.0198, 0.0108, 0.0135], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:26:18,437 INFO [train.py:901] (1/2) Epoch 6, batch 450, loss[loss=0.2247, simple_loss=0.2761, pruned_loss=0.08661, over 7317.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.2827, pruned_loss=0.08408, over 1294801.31 frames. ], batch size: 59, lr: 2.50e-02, grad_scale: 16.0 +2023-03-20 19:26:22,556 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 19:26:23,023 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 19:26:37,408 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.357e+02 3.392e+02 4.278e+02 5.459e+02 9.889e+02, threshold=8.555e+02, percent-clipped=5.0 +2023-03-20 19:26:45,003 INFO [train.py:901] (1/2) Epoch 6, batch 500, loss[loss=0.2089, simple_loss=0.2743, pruned_loss=0.07179, over 7322.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.2827, pruned_loss=0.08413, over 1329138.14 frames. ], batch size: 54, lr: 2.49e-02, grad_scale: 16.0 +2023-03-20 19:26:55,598 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 19:26:56,649 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 19:26:57,173 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 19:26:59,591 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 19:27:04,206 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 19:27:04,517 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-03-20 19:27:11,364 INFO [train.py:901] (1/2) Epoch 6, batch 550, loss[loss=0.2659, simple_loss=0.3158, pruned_loss=0.108, over 7159.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.2833, pruned_loss=0.08446, over 1353386.33 frames. ], batch size: 98, lr: 2.49e-02, grad_scale: 16.0 +2023-03-20 19:27:16,877 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 19:27:19,289 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 +2023-03-20 19:27:25,505 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 19:27:28,537 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 19:27:29,018 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.843e+02 3.123e+02 4.272e+02 5.651e+02 1.008e+03, threshold=8.544e+02, percent-clipped=3.0 +2023-03-20 19:27:36,277 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 19:27:36,785 INFO [train.py:901] (1/2) Epoch 6, batch 600, loss[loss=0.215, simple_loss=0.2707, pruned_loss=0.0796, over 7351.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.2822, pruned_loss=0.08477, over 1367743.16 frames. ], batch size: 44, lr: 2.49e-02, grad_scale: 16.0 +2023-03-20 19:27:40,453 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14728.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:27:46,563 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14740.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:27:52,409 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 19:27:56,591 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14759.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:28:02,296 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14769.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:28:03,292 INFO [train.py:901] (1/2) Epoch 6, batch 650, loss[loss=0.1742, simple_loss=0.2248, pruned_loss=0.06181, over 6465.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.2809, pruned_loss=0.08372, over 1379423.65 frames. ], batch size: 28, lr: 2.48e-02, grad_scale: 16.0 +2023-03-20 19:28:03,308 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 19:28:05,882 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=14776.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:28:05,952 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14776.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 19:28:18,879 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14801.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:28:20,831 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8722, 3.4987, 3.5390, 3.5313, 3.5619, 3.6184, 3.8183, 3.6304], + device='cuda:1'), covar=tensor([0.0084, 0.0153, 0.0146, 0.0188, 0.0166, 0.0120, 0.0147, 0.0110], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0044, 0.0049, 0.0042, 0.0059, 0.0052, 0.0046, 0.0043], + device='cuda:1'), out_proj_covar=tensor([8.8582e-05, 1.1500e-04, 1.2129e-04, 1.0506e-04, 1.5558e-04, 1.3722e-04, + 1.2740e-04, 1.0455e-04], device='cuda:1') +2023-03-20 19:28:21,200 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.066e+02 3.284e+02 4.078e+02 5.127e+02 1.452e+03, threshold=8.156e+02, percent-clipped=2.0 +2023-03-20 19:28:21,229 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 19:28:21,784 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=14807.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:28:28,645 INFO [train.py:901] (1/2) Epoch 6, batch 700, loss[loss=0.2538, simple_loss=0.298, pruned_loss=0.1048, over 7354.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.2811, pruned_loss=0.084, over 1392762.37 frames. ], batch size: 63, lr: 2.48e-02, grad_scale: 16.0 +2023-03-20 19:28:29,176 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 19:28:30,180 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=14824.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 19:28:39,583 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14843.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:28:53,193 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 19:28:53,723 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 19:28:54,713 INFO [train.py:901] (1/2) Epoch 6, batch 750, loss[loss=0.2434, simple_loss=0.304, pruned_loss=0.0914, over 7260.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.2825, pruned_loss=0.08485, over 1403493.95 frames. ], batch size: 55, lr: 2.47e-02, grad_scale: 16.0 +2023-03-20 19:29:04,513 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2852, 3.4625, 2.3754, 3.5088, 2.8552, 3.4481, 2.2356, 2.0711], + device='cuda:1'), covar=tensor([0.0023, 0.0118, 0.0534, 0.0070, 0.0055, 0.0060, 0.0675, 0.0472], + device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0141, 0.0257, 0.0119, 0.0132, 0.0131, 0.0251, 0.0246], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 19:29:07,878 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 19:29:11,926 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 19:29:12,401 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.820e+02 3.169e+02 4.239e+02 5.428e+02 1.155e+03, threshold=8.477e+02, percent-clipped=3.0 +2023-03-20 19:29:17,408 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 19:29:18,890 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 19:29:19,964 INFO [train.py:901] (1/2) Epoch 6, batch 800, loss[loss=0.216, simple_loss=0.2753, pruned_loss=0.07837, over 7254.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.2824, pruned_loss=0.08433, over 1412626.06 frames. ], batch size: 47, lr: 2.47e-02, grad_scale: 16.0 +2023-03-20 19:29:30,145 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 19:29:45,454 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2404, 4.0364, 3.5397, 4.1944, 3.1029, 2.8024, 4.0843, 3.3676], + device='cuda:1'), covar=tensor([0.0051, 0.0060, 0.0098, 0.0024, 0.0148, 0.0220, 0.0107, 0.0237], + device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0168, 0.0205, 0.0140, 0.0255, 0.0271, 0.0200, 0.0285], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 19:29:46,273 INFO [train.py:901] (1/2) Epoch 6, batch 850, loss[loss=0.2303, simple_loss=0.2946, pruned_loss=0.08298, over 7211.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.2828, pruned_loss=0.08419, over 1420161.10 frames. ], batch size: 50, lr: 2.47e-02, grad_scale: 16.0 +2023-03-20 19:29:46,928 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7318, 4.1692, 3.7393, 3.6869, 3.7584, 3.8154, 3.9725, 3.6318], + device='cuda:1'), covar=tensor([0.0067, 0.0059, 0.0069, 0.0079, 0.0067, 0.0062, 0.0061, 0.0094], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0030, 0.0030, 0.0027, 0.0028, 0.0028, 0.0035, 0.0033], + device='cuda:1'), out_proj_covar=tensor([8.3543e-05, 9.7505e-05, 1.0321e-04, 7.9559e-05, 8.9769e-05, 8.7949e-05, + 1.1501e-04, 1.0661e-04], device='cuda:1') +2023-03-20 19:29:48,328 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 19:29:48,341 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 19:29:53,974 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 19:29:57,547 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 19:30:04,431 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.946e+02 3.394e+02 4.189e+02 5.011e+02 1.297e+03, threshold=8.377e+02, percent-clipped=4.0 +2023-03-20 19:30:13,332 INFO [train.py:901] (1/2) Epoch 6, batch 900, loss[loss=0.2198, simple_loss=0.2807, pruned_loss=0.07943, over 7315.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.2825, pruned_loss=0.08425, over 1423267.06 frames. ], batch size: 49, lr: 2.46e-02, grad_scale: 16.0 +2023-03-20 19:30:36,463 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 19:30:37,560 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15069.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:30:38,482 INFO [train.py:901] (1/2) Epoch 6, batch 950, loss[loss=0.2337, simple_loss=0.2932, pruned_loss=0.08712, over 7343.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.2823, pruned_loss=0.08404, over 1428525.47 frames. ], batch size: 61, lr: 2.46e-02, grad_scale: 16.0 +2023-03-20 19:30:51,150 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15096.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:30:56,598 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.209e+02 3.393e+02 3.946e+02 5.314e+02 1.457e+03, threshold=7.892e+02, percent-clipped=4.0 +2023-03-20 19:31:00,375 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2493, 4.0086, 3.9490, 3.6722, 3.8336, 2.5326, 2.2404, 4.2960], + device='cuda:1'), covar=tensor([0.0017, 0.0038, 0.0059, 0.0060, 0.0069, 0.0363, 0.0620, 0.0039], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0050, 0.0068, 0.0052, 0.0056, 0.0082, 0.0097, 0.0055], + device='cuda:1'), out_proj_covar=tensor([6.3627e-05, 7.7333e-05, 9.9651e-05, 7.8120e-05, 7.4713e-05, 1.2134e-04, + 1.4431e-04, 7.6774e-05], device='cuda:1') +2023-03-20 19:31:00,777 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 19:31:02,820 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=15117.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:31:04,789 INFO [train.py:901] (1/2) Epoch 6, batch 1000, loss[loss=0.2412, simple_loss=0.2898, pruned_loss=0.09626, over 7314.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.2814, pruned_loss=0.08311, over 1432556.71 frames. ], batch size: 83, lr: 2.46e-02, grad_scale: 16.0 +2023-03-20 19:31:15,980 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15143.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:31:22,288 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 19:31:29,712 INFO [train.py:901] (1/2) Epoch 6, batch 1050, loss[loss=0.2107, simple_loss=0.2754, pruned_loss=0.07301, over 7307.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.2819, pruned_loss=0.08348, over 1431685.55 frames. ], batch size: 80, lr: 2.45e-02, grad_scale: 16.0 +2023-03-20 19:31:37,928 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9496, 3.5851, 2.9635, 3.8907, 2.7214, 2.5676, 3.9520, 2.9903], + device='cuda:1'), covar=tensor([0.0053, 0.0053, 0.0201, 0.0027, 0.0236, 0.0303, 0.0101, 0.0420], + device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0172, 0.0208, 0.0141, 0.0258, 0.0274, 0.0203, 0.0283], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 19:31:39,835 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=15191.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:31:42,603 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15195.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:31:43,493 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 19:31:48,945 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.607e+02 3.249e+02 4.260e+02 5.162e+02 1.307e+03, threshold=8.520e+02, percent-clipped=8.0 +2023-03-20 19:31:48,981 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 19:31:56,449 INFO [train.py:901] (1/2) Epoch 6, batch 1100, loss[loss=0.2232, simple_loss=0.278, pruned_loss=0.08421, over 7274.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.2835, pruned_loss=0.08389, over 1436186.49 frames. ], batch size: 70, lr: 2.45e-02, grad_scale: 16.0 +2023-03-20 19:32:13,952 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.00 vs. limit=2.0 +2023-03-20 19:32:14,290 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15256.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 19:32:17,130 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 19:32:17,621 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 19:32:21,571 INFO [train.py:901] (1/2) Epoch 6, batch 1150, loss[loss=0.2385, simple_loss=0.2938, pruned_loss=0.09154, over 7278.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.2833, pruned_loss=0.08386, over 1436127.82 frames. ], batch size: 47, lr: 2.44e-02, grad_scale: 16.0 +2023-03-20 19:32:21,745 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3419, 2.2112, 2.1890, 3.2642, 2.8206, 2.9777, 2.9694, 2.8829], + device='cuda:1'), covar=tensor([0.0715, 0.0323, 0.1014, 0.0261, 0.0014, 0.0025, 0.0054, 0.0021], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0191, 0.0238, 0.0159, 0.0096, 0.0102, 0.0098, 0.0097], + device='cuda:1'), out_proj_covar=tensor([2.3684e-04, 1.8585e-04, 2.2051e-04, 1.5885e-04, 9.4720e-05, 1.0272e-04, + 9.6116e-05, 9.9692e-05], device='cuda:1') +2023-03-20 19:32:30,778 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 19:32:31,758 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 19:32:40,305 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.354e+02 3.501e+02 4.043e+02 4.936e+02 1.554e+03, threshold=8.087e+02, percent-clipped=2.0 +2023-03-20 19:32:47,945 INFO [train.py:901] (1/2) Epoch 6, batch 1200, loss[loss=0.2112, simple_loss=0.2728, pruned_loss=0.07478, over 7277.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.2819, pruned_loss=0.08334, over 1436518.12 frames. ], batch size: 70, lr: 2.44e-02, grad_scale: 16.0 +2023-03-20 19:32:49,735 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.16 vs. limit=2.0 +2023-03-20 19:32:57,132 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2561, 3.4214, 2.5344, 3.7651, 2.9508, 3.3174, 2.2472, 2.2815], + device='cuda:1'), covar=tensor([0.0037, 0.0231, 0.0592, 0.0076, 0.0064, 0.0098, 0.0984, 0.0611], + device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0142, 0.0243, 0.0121, 0.0128, 0.0135, 0.0237, 0.0232], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 19:33:03,384 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 19:33:13,668 INFO [train.py:901] (1/2) Epoch 6, batch 1250, loss[loss=0.2091, simple_loss=0.2798, pruned_loss=0.06915, over 7309.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.2816, pruned_loss=0.08281, over 1440218.77 frames. ], batch size: 80, lr: 2.44e-02, grad_scale: 16.0 +2023-03-20 19:33:25,069 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1072, 2.5831, 2.6287, 2.2670, 2.1509, 2.2846, 1.8073, 2.2272], + device='cuda:1'), covar=tensor([0.1482, 0.0480, 0.1906, 0.2170, 0.1585, 0.1985, 0.2633, 0.1452], + device='cuda:1'), in_proj_covar=tensor([0.0026, 0.0026, 0.0027, 0.0029, 0.0025, 0.0026, 0.0036, 0.0027], + device='cuda:1'), out_proj_covar=tensor([7.3714e-05, 6.6732e-05, 7.5384e-05, 7.9625e-05, 7.0296e-05, 7.5491e-05, + 9.0891e-05, 7.4611e-05], device='cuda:1') +2023-03-20 19:33:26,903 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 19:33:27,035 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15396.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:33:30,229 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-03-20 19:33:30,803 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 19:33:32,288 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.950e+02 3.210e+02 3.988e+02 5.290e+02 1.673e+03, threshold=7.977e+02, percent-clipped=3.0 +2023-03-20 19:33:32,326 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 19:33:39,846 INFO [train.py:901] (1/2) Epoch 6, batch 1300, loss[loss=0.2184, simple_loss=0.2759, pruned_loss=0.08043, over 7225.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.2814, pruned_loss=0.08291, over 1438641.09 frames. ], batch size: 45, lr: 2.43e-02, grad_scale: 16.0 +2023-03-20 19:33:40,177 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.20 vs. limit=2.0 +2023-03-20 19:33:51,382 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=15444.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:33:54,840 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 19:33:57,398 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 19:34:01,570 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 19:34:06,036 INFO [train.py:901] (1/2) Epoch 6, batch 1350, loss[loss=0.2246, simple_loss=0.28, pruned_loss=0.08461, over 7334.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.2813, pruned_loss=0.08278, over 1441211.72 frames. ], batch size: 54, lr: 2.43e-02, grad_scale: 16.0 +2023-03-20 19:34:12,056 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 19:34:12,877 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-03-20 19:34:15,194 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2896, 4.1159, 4.2192, 4.4829, 4.6536, 4.6432, 3.7593, 4.0259], + device='cuda:1'), covar=tensor([0.0851, 0.2445, 0.2290, 0.1207, 0.0636, 0.1189, 0.0936, 0.0994], + device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0196, 0.0186, 0.0163, 0.0133, 0.0203, 0.0115, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:34:23,684 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.201e+02 3.158e+02 3.725e+02 4.655e+02 8.945e+02, threshold=7.450e+02, percent-clipped=1.0 +2023-03-20 19:34:31,276 INFO [train.py:901] (1/2) Epoch 6, batch 1400, loss[loss=0.2622, simple_loss=0.3055, pruned_loss=0.1094, over 7323.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.2807, pruned_loss=0.0823, over 1441204.93 frames. ], batch size: 54, lr: 2.43e-02, grad_scale: 16.0 +2023-03-20 19:34:35,274 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 +2023-03-20 19:34:45,826 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 19:34:47,829 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15551.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 19:34:56,565 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.11 vs. limit=2.0 +2023-03-20 19:34:57,764 INFO [train.py:901] (1/2) Epoch 6, batch 1450, loss[loss=0.2443, simple_loss=0.3044, pruned_loss=0.09213, over 7324.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.2805, pruned_loss=0.08238, over 1440304.42 frames. ], batch size: 61, lr: 2.42e-02, grad_scale: 16.0 +2023-03-20 19:35:08,882 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 19:35:15,839 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.277e+02 3.171e+02 3.854e+02 4.989e+02 1.002e+03, threshold=7.708e+02, percent-clipped=5.0 +2023-03-20 19:35:24,023 INFO [train.py:901] (1/2) Epoch 6, batch 1500, loss[loss=0.2328, simple_loss=0.2903, pruned_loss=0.08768, over 7348.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.2806, pruned_loss=0.08224, over 1441662.60 frames. ], batch size: 54, lr: 2.42e-02, grad_scale: 16.0 +2023-03-20 19:35:25,033 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 19:35:36,742 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4699, 3.1439, 3.1781, 3.1993, 3.0613, 3.0764, 3.2517, 3.0585], + device='cuda:1'), covar=tensor([0.0153, 0.0256, 0.0242, 0.0277, 0.0293, 0.0221, 0.0338, 0.0264], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0044, 0.0048, 0.0042, 0.0058, 0.0051, 0.0047, 0.0045], + device='cuda:1'), out_proj_covar=tensor([9.1774e-05, 1.1381e-04, 1.2074e-04, 1.0594e-04, 1.5029e-04, 1.3317e-04, + 1.3082e-04, 1.0907e-04], device='cuda:1') +2023-03-20 19:35:39,215 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0262, 2.4130, 2.7458, 2.2862, 2.1984, 2.2765, 1.7073, 2.0931], + device='cuda:1'), covar=tensor([0.1355, 0.0179, 0.0793, 0.2296, 0.0954, 0.1740, 0.2795, 0.1604], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0023, 0.0025, 0.0028, 0.0023, 0.0024, 0.0033, 0.0025], + device='cuda:1'), out_proj_covar=tensor([7.0638e-05, 6.1999e-05, 7.1645e-05, 7.6266e-05, 6.7018e-05, 7.1853e-05, + 8.5078e-05, 7.0562e-05], device='cuda:1') +2023-03-20 19:35:48,806 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 19:35:49,785 INFO [train.py:901] (1/2) Epoch 6, batch 1550, loss[loss=0.2008, simple_loss=0.2662, pruned_loss=0.06769, over 7216.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.2799, pruned_loss=0.08169, over 1440578.94 frames. ], batch size: 50, lr: 2.42e-02, grad_scale: 16.0 +2023-03-20 19:35:58,098 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.16 vs. limit=5.0 +2023-03-20 19:36:02,152 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.08 vs. limit=2.0 +2023-03-20 19:36:07,855 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.699e+02 3.513e+02 4.071e+02 5.180e+02 1.061e+03, threshold=8.142e+02, percent-clipped=4.0 +2023-03-20 19:36:14,874 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-03-20 19:36:15,198 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.7166, 1.0327, 1.0298, 1.1841, 0.8511, 0.9583, 0.9301, 0.9760], + device='cuda:1'), covar=tensor([0.1221, 0.1030, 0.0389, 0.0502, 0.0931, 0.1179, 0.0361, 0.0689], + device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0035, 0.0031, 0.0029, 0.0032, 0.0030, 0.0034, 0.0032], + device='cuda:1'), out_proj_covar=tensor([5.6023e-05, 6.9965e-05, 4.7006e-05, 4.5714e-05, 5.4372e-05, 5.4556e-05, + 5.6755e-05, 5.7426e-05], device='cuda:1') +2023-03-20 19:36:16,047 INFO [train.py:901] (1/2) Epoch 6, batch 1600, loss[loss=0.2427, simple_loss=0.2989, pruned_loss=0.09328, over 7261.00 frames. ], tot_loss[loss=0.222, simple_loss=0.2799, pruned_loss=0.08202, over 1439366.33 frames. ], batch size: 89, lr: 2.41e-02, grad_scale: 16.0 +2023-03-20 19:36:20,516 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 19:36:21,009 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 19:36:23,662 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9352, 4.2497, 3.8044, 4.0130, 3.8870, 4.1349, 4.2410, 4.3978], + device='cuda:1'), covar=tensor([0.0325, 0.0210, 0.0357, 0.0330, 0.0484, 0.0213, 0.0423, 0.0292], + device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0084, 0.0081, 0.0091, 0.0090, 0.0068, 0.0069, 0.0069], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:36:24,572 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 19:36:34,097 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 19:36:38,740 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 19:36:41,167 INFO [train.py:901] (1/2) Epoch 6, batch 1650, loss[loss=0.2198, simple_loss=0.2817, pruned_loss=0.0789, over 7315.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.2777, pruned_loss=0.08088, over 1438381.88 frames. ], batch size: 49, lr: 2.41e-02, grad_scale: 16.0 +2023-03-20 19:36:46,675 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 19:36:59,798 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.745e+02 3.076e+02 3.798e+02 5.090e+02 1.024e+03, threshold=7.597e+02, percent-clipped=1.0 +2023-03-20 19:37:05,421 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 19:37:07,891 INFO [train.py:901] (1/2) Epoch 6, batch 1700, loss[loss=0.2027, simple_loss=0.2589, pruned_loss=0.0733, over 7228.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2785, pruned_loss=0.08134, over 1439599.65 frames. ], batch size: 45, lr: 2.41e-02, grad_scale: 16.0 +2023-03-20 19:37:09,409 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 19:37:19,446 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 19:37:23,037 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15851.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:37:32,960 INFO [train.py:901] (1/2) Epoch 6, batch 1750, loss[loss=0.1982, simple_loss=0.2666, pruned_loss=0.06493, over 7336.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.278, pruned_loss=0.08104, over 1439403.21 frames. ], batch size: 75, lr: 2.40e-02, grad_scale: 32.0 +2023-03-20 19:37:39,755 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15883.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:37:43,120 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 19:37:44,097 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 19:37:48,243 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=15899.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:37:51,730 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.174e+02 3.400e+02 4.133e+02 5.525e+02 1.160e+03, threshold=8.266e+02, percent-clipped=5.0 +2023-03-20 19:37:59,239 INFO [train.py:901] (1/2) Epoch 6, batch 1800, loss[loss=0.1751, simple_loss=0.2237, pruned_loss=0.06327, over 6253.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2789, pruned_loss=0.08163, over 1439726.34 frames. ], batch size: 27, lr: 2.40e-02, grad_scale: 32.0 +2023-03-20 19:38:05,879 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 19:38:11,002 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15944.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:38:19,485 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 19:38:25,097 INFO [train.py:901] (1/2) Epoch 6, batch 1850, loss[loss=0.2441, simple_loss=0.3048, pruned_loss=0.09168, over 7343.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2785, pruned_loss=0.08121, over 1439782.73 frames. ], batch size: 63, lr: 2.40e-02, grad_scale: 32.0 +2023-03-20 19:38:30,069 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 19:38:44,653 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16001.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:38:46,989 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 3.006e+02 3.753e+02 4.751e+02 8.909e+02, threshold=7.505e+02, percent-clipped=1.0 +2023-03-20 19:38:50,559 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 19:38:50,862 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.17 vs. limit=2.0 +2023-03-20 19:38:51,697 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0854, 3.4513, 2.3986, 3.5912, 2.8942, 3.3025, 2.4704, 1.9527], + device='cuda:1'), covar=tensor([0.0019, 0.0176, 0.0461, 0.0055, 0.0074, 0.0095, 0.0603, 0.0492], + device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0149, 0.0257, 0.0134, 0.0145, 0.0141, 0.0242, 0.0241], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 19:38:52,644 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.9805, 1.8465, 1.4046, 1.7012, 0.8554, 1.4381, 1.1226, 1.2231], + device='cuda:1'), covar=tensor([0.0556, 0.0208, 0.0265, 0.0109, 0.0517, 0.0217, 0.0459, 0.0340], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0017, 0.0017, 0.0018, 0.0020, 0.0016, 0.0019, 0.0020], + device='cuda:1'), out_proj_covar=tensor([4.2475e-05, 3.7561e-05, 3.3742e-05, 3.5349e-05, 4.3431e-05, 3.6831e-05, + 4.1730e-05, 4.4278e-05], device='cuda:1') +2023-03-20 19:38:54,479 INFO [train.py:901] (1/2) Epoch 6, batch 1900, loss[loss=0.2276, simple_loss=0.2946, pruned_loss=0.0803, over 7122.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.2794, pruned_loss=0.08147, over 1442545.05 frames. ], batch size: 98, lr: 2.39e-02, grad_scale: 32.0 +2023-03-20 19:39:02,935 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-03-20 19:39:15,734 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 19:39:15,866 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16062.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:39:20,791 INFO [train.py:901] (1/2) Epoch 6, batch 1950, loss[loss=0.2332, simple_loss=0.2893, pruned_loss=0.08856, over 7296.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.2805, pruned_loss=0.08195, over 1445602.62 frames. ], batch size: 59, lr: 2.39e-02, grad_scale: 16.0 +2023-03-20 19:39:23,610 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-20 19:39:27,834 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 19:39:27,959 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16085.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:39:32,277 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 19:39:32,767 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 19:39:38,698 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.218e+02 3.583e+02 4.318e+02 5.667e+02 1.031e+03, threshold=8.635e+02, percent-clipped=6.0 +2023-03-20 19:39:45,755 INFO [train.py:901] (1/2) Epoch 6, batch 2000, loss[loss=0.1811, simple_loss=0.2368, pruned_loss=0.06268, over 7012.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.2805, pruned_loss=0.0821, over 1446540.58 frames. ], batch size: 35, lr: 2.39e-02, grad_scale: 16.0 +2023-03-20 19:39:48,289 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 19:39:59,101 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16146.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:39:59,967 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 19:40:00,613 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16149.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:40:09,671 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 19:40:12,235 INFO [train.py:901] (1/2) Epoch 6, batch 2050, loss[loss=0.2066, simple_loss=0.271, pruned_loss=0.07106, over 7354.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2794, pruned_loss=0.08124, over 1444168.65 frames. ], batch size: 73, lr: 2.38e-02, grad_scale: 16.0 +2023-03-20 19:40:30,496 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 3.318e+02 4.048e+02 5.137e+02 9.969e+02, threshold=8.096e+02, percent-clipped=1.0 +2023-03-20 19:40:32,180 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16210.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:40:37,590 INFO [train.py:901] (1/2) Epoch 6, batch 2100, loss[loss=0.2418, simple_loss=0.2965, pruned_loss=0.09358, over 7261.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2792, pruned_loss=0.08125, over 1443364.74 frames. ], batch size: 52, lr: 2.38e-02, grad_scale: 16.0 +2023-03-20 19:40:42,797 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 19:40:45,302 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 19:40:47,408 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16239.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:41:04,147 INFO [train.py:901] (1/2) Epoch 6, batch 2150, loss[loss=0.201, simple_loss=0.2636, pruned_loss=0.06919, over 7323.00 frames. ], tot_loss[loss=0.22, simple_loss=0.2787, pruned_loss=0.08069, over 1443328.31 frames. ], batch size: 75, lr: 2.38e-02, grad_scale: 16.0 +2023-03-20 19:41:22,166 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.745e+02 3.153e+02 4.020e+02 4.904e+02 9.769e+02, threshold=8.040e+02, percent-clipped=2.0 +2023-03-20 19:41:29,916 INFO [train.py:901] (1/2) Epoch 6, batch 2200, loss[loss=0.2231, simple_loss=0.2776, pruned_loss=0.08429, over 7355.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2781, pruned_loss=0.07996, over 1442641.56 frames. ], batch size: 51, lr: 2.37e-02, grad_scale: 16.0 +2023-03-20 19:41:33,952 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 19:41:48,544 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16357.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:41:53,798 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.04 vs. limit=2.0 +2023-03-20 19:41:55,490 INFO [train.py:901] (1/2) Epoch 6, batch 2250, loss[loss=0.2337, simple_loss=0.2866, pruned_loss=0.09038, over 7254.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2784, pruned_loss=0.08038, over 1444196.06 frames. ], batch size: 55, lr: 2.37e-02, grad_scale: 16.0 +2023-03-20 19:41:55,595 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16371.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:42:06,629 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2012, 1.7329, 1.5597, 1.3285, 1.0635, 1.7572, 1.3762, 1.1938], + device='cuda:1'), covar=tensor([0.0451, 0.0232, 0.0167, 0.0133, 0.0377, 0.0213, 0.0163, 0.0277], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0019, 0.0017, 0.0019, 0.0019, 0.0016, 0.0020, 0.0020], + device='cuda:1'), out_proj_covar=tensor([4.4018e-05, 4.0937e-05, 3.5563e-05, 3.7205e-05, 4.3161e-05, 3.7690e-05, + 4.3130e-05, 4.6424e-05], device='cuda:1') +2023-03-20 19:42:07,540 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 19:42:08,018 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 19:42:13,261 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.8414, 0.9726, 0.8402, 1.2328, 0.8824, 0.9831, 1.1849, 1.1225], + device='cuda:1'), covar=tensor([0.1012, 0.0883, 0.0373, 0.0163, 0.0801, 0.0520, 0.0175, 0.0453], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0034, 0.0029, 0.0027, 0.0029, 0.0026, 0.0029, 0.0029], + device='cuda:1'), out_proj_covar=tensor([5.2719e-05, 6.7329e-05, 4.5984e-05, 4.3610e-05, 5.0962e-05, 4.8702e-05, + 4.9331e-05, 5.4048e-05], device='cuda:1') +2023-03-20 19:42:14,075 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.726e+02 3.600e+02 4.424e+02 5.691e+02 1.633e+03, threshold=8.848e+02, percent-clipped=4.0 +2023-03-20 19:42:18,471 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.61 vs. limit=5.0 +2023-03-20 19:42:20,691 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 19:42:21,194 INFO [train.py:901] (1/2) Epoch 6, batch 2300, loss[loss=0.1929, simple_loss=0.2645, pruned_loss=0.06067, over 7350.00 frames. ], tot_loss[loss=0.22, simple_loss=0.2787, pruned_loss=0.08061, over 1443080.74 frames. ], batch size: 73, lr: 2.37e-02, grad_scale: 16.0 +2023-03-20 19:42:27,544 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16432.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:42:31,102 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2621, 2.4528, 2.6560, 3.3678, 3.3633, 3.5502, 3.3478, 3.0181], + device='cuda:1'), covar=tensor([0.0969, 0.0429, 0.1125, 0.0240, 0.0045, 0.0048, 0.0044, 0.0050], + device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0202, 0.0256, 0.0179, 0.0104, 0.0101, 0.0103, 0.0109], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:42:32,021 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16441.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:42:32,531 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6045, 4.5222, 4.6824, 4.9945, 5.1373, 5.0577, 4.3321, 4.5565], + device='cuda:1'), covar=tensor([0.0792, 0.2187, 0.1814, 0.1082, 0.0490, 0.1194, 0.0640, 0.0915], + device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0204, 0.0189, 0.0170, 0.0136, 0.0221, 0.0120, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:42:34,081 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8216, 1.9286, 1.8936, 2.0438, 1.9257, 1.9728, 2.0886, 2.0849], + device='cuda:1'), covar=tensor([0.0628, 0.0709, 0.0600, 0.0528, 0.0812, 0.0402, 0.0309, 0.0432], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0035, 0.0035, 0.0035, 0.0034, 0.0035, 0.0033, 0.0034], + device='cuda:1'), out_proj_covar=tensor([9.5043e-05, 8.6949e-05, 8.5758e-05, 8.5367e-05, 8.6198e-05, 8.7720e-05, + 8.3238e-05, 8.5520e-05], device='cuda:1') +2023-03-20 19:42:42,507 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0128, 3.3280, 3.1213, 2.8120, 3.4205, 2.9549, 2.5380, 2.7842], + device='cuda:1'), covar=tensor([0.1295, 0.0207, 0.1090, 0.3045, 0.0363, 0.2148, 0.2404, 0.1834], + device='cuda:1'), in_proj_covar=tensor([0.0026, 0.0024, 0.0028, 0.0029, 0.0025, 0.0027, 0.0033, 0.0027], + device='cuda:1'), out_proj_covar=tensor([7.6691e-05, 6.7808e-05, 7.9982e-05, 8.2776e-05, 7.2458e-05, 7.9072e-05, + 9.1203e-05, 7.8426e-05], device='cuda:1') +2023-03-20 19:42:44,989 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9070, 3.8825, 3.5514, 3.7546, 3.5277, 4.0485, 4.3442, 4.3158], + device='cuda:1'), covar=tensor([0.0195, 0.0203, 0.0294, 0.0226, 0.0497, 0.0222, 0.0237, 0.0229], + device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0089, 0.0082, 0.0094, 0.0088, 0.0070, 0.0068, 0.0071], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-03-20 19:42:46,892 INFO [train.py:901] (1/2) Epoch 6, batch 2350, loss[loss=0.2458, simple_loss=0.3057, pruned_loss=0.09288, over 7142.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2798, pruned_loss=0.0807, over 1445188.55 frames. ], batch size: 98, lr: 2.36e-02, grad_scale: 16.0 +2023-03-20 19:43:04,581 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16505.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:43:05,504 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.961e+02 3.138e+02 3.927e+02 4.655e+02 8.908e+02, threshold=7.854e+02, percent-clipped=1.0 +2023-03-20 19:43:06,570 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 19:43:06,698 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6180, 2.6275, 2.5519, 2.6637, 2.9782, 2.7529, 1.8711, 2.2485], + device='cuda:1'), covar=tensor([0.0982, 0.0198, 0.0856, 0.1722, 0.0428, 0.0843, 0.2596, 0.1770], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0025, 0.0027, 0.0029, 0.0024, 0.0026, 0.0034, 0.0027], + device='cuda:1'), out_proj_covar=tensor([7.7994e-05, 6.9305e-05, 7.9783e-05, 8.2879e-05, 7.1712e-05, 7.8260e-05, + 9.1625e-05, 7.9867e-05], device='cuda:1') +2023-03-20 19:43:10,374 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8757, 3.2077, 3.4905, 3.4407, 3.5067, 3.5674, 3.6865, 3.2425], + device='cuda:1'), covar=tensor([0.0108, 0.0242, 0.0189, 0.0171, 0.0187, 0.0116, 0.0159, 0.0175], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0042, 0.0050, 0.0042, 0.0058, 0.0052, 0.0047, 0.0045], + device='cuda:1'), out_proj_covar=tensor([9.3927e-05, 1.1231e-04, 1.2524e-04, 1.0575e-04, 1.5291e-04, 1.3860e-04, + 1.3158e-04, 1.0891e-04], device='cuda:1') +2023-03-20 19:43:13,371 INFO [train.py:901] (1/2) Epoch 6, batch 2400, loss[loss=0.2149, simple_loss=0.2779, pruned_loss=0.076, over 7296.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2803, pruned_loss=0.08065, over 1445328.87 frames. ], batch size: 86, lr: 2.36e-02, grad_scale: 16.0 +2023-03-20 19:43:13,938 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 19:43:22,488 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 19:43:22,583 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16539.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:43:25,392 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 19:43:39,045 INFO [train.py:901] (1/2) Epoch 6, batch 2450, loss[loss=0.2277, simple_loss=0.2905, pruned_loss=0.08244, over 7274.00 frames. ], tot_loss[loss=0.222, simple_loss=0.2814, pruned_loss=0.08127, over 1445409.65 frames. ], batch size: 52, lr: 2.36e-02, grad_scale: 16.0 +2023-03-20 19:43:47,153 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=16587.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:43:52,655 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 19:43:58,217 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 3.302e+02 4.000e+02 4.919e+02 9.001e+02, threshold=8.000e+02, percent-clipped=2.0 +2023-03-20 19:44:05,296 INFO [train.py:901] (1/2) Epoch 6, batch 2500, loss[loss=0.2112, simple_loss=0.2735, pruned_loss=0.07447, over 7355.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2795, pruned_loss=0.08082, over 1444067.13 frames. ], batch size: 63, lr: 2.35e-02, grad_scale: 16.0 +2023-03-20 19:44:18,914 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 19:44:21,727 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 +2023-03-20 19:44:24,197 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16657.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:44:31,138 INFO [train.py:901] (1/2) Epoch 6, batch 2550, loss[loss=0.2035, simple_loss=0.2635, pruned_loss=0.07173, over 7343.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.2806, pruned_loss=0.08137, over 1443200.44 frames. ], batch size: 44, lr: 2.35e-02, grad_scale: 16.0 +2023-03-20 19:44:49,000 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=16705.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:44:49,938 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.726e+02 3.249e+02 3.789e+02 4.913e+02 1.049e+03, threshold=7.579e+02, percent-clipped=3.0 +2023-03-20 19:44:56,969 INFO [train.py:901] (1/2) Epoch 6, batch 2600, loss[loss=0.2205, simple_loss=0.2837, pruned_loss=0.07868, over 7321.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.28, pruned_loss=0.08107, over 1439749.18 frames. ], batch size: 83, lr: 2.35e-02, grad_scale: 16.0 +2023-03-20 19:44:59,914 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16727.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:45:06,792 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16741.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:45:21,579 INFO [train.py:901] (1/2) Epoch 6, batch 2650, loss[loss=0.2193, simple_loss=0.2795, pruned_loss=0.07957, over 7259.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2796, pruned_loss=0.08079, over 1439780.39 frames. ], batch size: 52, lr: 2.34e-02, grad_scale: 16.0 +2023-03-20 19:45:21,702 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0742, 4.0067, 3.4972, 3.2671, 3.7071, 2.2552, 1.6355, 4.0621], + device='cuda:1'), covar=tensor([0.0011, 0.0022, 0.0073, 0.0078, 0.0034, 0.0414, 0.0783, 0.0033], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0050, 0.0072, 0.0055, 0.0057, 0.0086, 0.0098, 0.0057], + device='cuda:1'), out_proj_covar=tensor([6.5550e-05, 7.9493e-05, 1.0431e-04, 8.3314e-05, 7.5802e-05, 1.2655e-04, + 1.4530e-04, 8.1453e-05], device='cuda:1') +2023-03-20 19:45:30,427 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=16789.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:45:39,198 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16805.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:45:40,047 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.977e+02 3.511e+02 4.230e+02 5.351e+02 9.415e+02, threshold=8.460e+02, percent-clipped=2.0 +2023-03-20 19:45:43,148 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4187, 2.0561, 2.6604, 2.2144, 2.3138, 2.3604, 2.2008, 2.2906], + device='cuda:1'), covar=tensor([0.1413, 0.0285, 0.1036, 0.2540, 0.0705, 0.1613, 0.2012, 0.1430], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0024, 0.0027, 0.0028, 0.0024, 0.0024, 0.0034, 0.0026], + device='cuda:1'), out_proj_covar=tensor([7.4696e-05, 6.8003e-05, 7.8464e-05, 8.1880e-05, 7.1409e-05, 7.5586e-05, + 9.2348e-05, 7.7856e-05], device='cuda:1') +2023-03-20 19:45:46,940 INFO [train.py:901] (1/2) Epoch 6, batch 2700, loss[loss=0.2134, simple_loss=0.2753, pruned_loss=0.07581, over 7325.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.2802, pruned_loss=0.08121, over 1442451.82 frames. ], batch size: 54, lr: 2.34e-02, grad_scale: 16.0 +2023-03-20 19:46:02,718 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=16853.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:46:10,315 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16868.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 19:46:11,759 INFO [train.py:901] (1/2) Epoch 6, batch 2750, loss[loss=0.2276, simple_loss=0.295, pruned_loss=0.08014, over 7318.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2799, pruned_loss=0.08105, over 1442433.77 frames. ], batch size: 75, lr: 2.34e-02, grad_scale: 16.0 +2023-03-20 19:46:14,842 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7080, 4.0002, 3.8182, 3.7864, 3.6679, 3.7289, 3.9601, 3.9296], + device='cuda:1'), covar=tensor([0.0400, 0.0218, 0.0313, 0.0278, 0.0508, 0.0299, 0.0513, 0.0473], + device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0090, 0.0086, 0.0096, 0.0092, 0.0069, 0.0071, 0.0073], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 19:46:27,931 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5294, 4.9795, 5.0205, 4.8816, 4.6573, 4.4518, 4.9928, 4.7809], + device='cuda:1'), covar=tensor([0.0342, 0.0319, 0.0371, 0.0443, 0.0335, 0.0323, 0.0309, 0.0446], + device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0124, 0.0105, 0.0095, 0.0087, 0.0119, 0.0109, 0.0087], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 19:46:28,953 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16906.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:46:29,263 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.247e+02 3.398e+02 4.341e+02 5.354e+02 9.122e+02, threshold=8.682e+02, percent-clipped=2.0 +2023-03-20 19:46:36,490 INFO [train.py:901] (1/2) Epoch 6, batch 2800, loss[loss=0.2339, simple_loss=0.2972, pruned_loss=0.08527, over 7226.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.2803, pruned_loss=0.08166, over 1439117.93 frames. ], batch size: 93, lr: 2.33e-02, grad_scale: 16.0 +2023-03-20 19:46:40,555 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16929.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 19:47:01,622 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 19:47:01,676 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 19:47:01,699 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 19:47:01,717 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 19:47:01,891 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 19:47:01,906 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 19:47:01,917 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 19:47:02,044 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 19:47:02,445 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 19:47:02,460 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 19:47:02,476 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 19:47:02,637 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 19:47:02,643 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 19:47:02,650 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 19:47:02,786 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 19:47:02,788 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 19:47:02,854 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 19:47:10,290 INFO [train.py:901] (1/2) Epoch 7, batch 0, loss[loss=0.2087, simple_loss=0.2839, pruned_loss=0.06674, over 7251.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2839, pruned_loss=0.06674, over 7251.00 frames. ], batch size: 55, lr: 2.24e-02, grad_scale: 16.0 +2023-03-20 19:47:10,291 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 19:47:26,150 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0470, 1.3534, 1.5312, 1.2191, 1.1676, 1.4930, 1.0191, 1.2568], + device='cuda:1'), covar=tensor([0.0481, 0.0332, 0.0191, 0.0152, 0.0226, 0.0325, 0.0393, 0.0549], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0020, 0.0020, 0.0020, 0.0021, 0.0019, 0.0020, 0.0023], + device='cuda:1'), out_proj_covar=tensor([4.8013e-05, 4.3185e-05, 4.0949e-05, 3.9016e-05, 4.7489e-05, 4.3493e-05, + 4.5166e-05, 5.0997e-05], device='cuda:1') +2023-03-20 19:47:36,469 INFO [train.py:935] (1/2) Epoch 7, validation: loss=0.1866, simple_loss=0.272, pruned_loss=0.05057, over 1622729.00 frames. +2023-03-20 19:47:36,470 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-20 19:47:43,610 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 19:47:47,770 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16967.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 19:47:54,613 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 19:47:58,085 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 +2023-03-20 19:48:01,257 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 19:48:01,728 INFO [train.py:901] (1/2) Epoch 7, batch 50, loss[loss=0.2417, simple_loss=0.2941, pruned_loss=0.09468, over 7316.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2772, pruned_loss=0.07933, over 327111.23 frames. ], batch size: 80, lr: 2.23e-02, grad_scale: 16.0 +2023-03-20 19:48:03,638 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 19:48:03,736 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9781, 4.1938, 4.0526, 4.0874, 3.7700, 4.2115, 4.5001, 4.4251], + device='cuda:1'), covar=tensor([0.0211, 0.0128, 0.0166, 0.0174, 0.0325, 0.0159, 0.0190, 0.0187], + device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0090, 0.0086, 0.0097, 0.0091, 0.0072, 0.0072, 0.0074], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 19:48:07,126 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 19:48:08,750 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.771e+02 3.154e+02 3.992e+02 5.174e+02 1.123e+03, threshold=7.983e+02, percent-clipped=2.0 +2023-03-20 19:48:10,168 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.04 vs. limit=2.0 +2023-03-20 19:48:18,894 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17027.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:48:20,629 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2477, 4.2493, 3.8034, 3.9385, 2.9958, 2.8065, 4.2050, 3.4321], + device='cuda:1'), covar=tensor([0.0138, 0.0060, 0.0087, 0.0037, 0.0196, 0.0264, 0.0114, 0.0277], + device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0186, 0.0209, 0.0154, 0.0265, 0.0276, 0.0203, 0.0282], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 19:48:21,742 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 19:48:25,424 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 19:48:25,892 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 19:48:28,411 INFO [train.py:901] (1/2) Epoch 7, batch 100, loss[loss=0.1988, simple_loss=0.2693, pruned_loss=0.06411, over 7317.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2769, pruned_loss=0.07919, over 570723.75 frames. ], batch size: 59, lr: 2.23e-02, grad_scale: 16.0 +2023-03-20 19:48:37,011 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3893, 1.8326, 1.5117, 1.2331, 1.5515, 1.5621, 1.3014, 1.2654], + device='cuda:1'), covar=tensor([0.0416, 0.0276, 0.0303, 0.0113, 0.0259, 0.0277, 0.0402, 0.0336], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0020, 0.0020, 0.0020, 0.0021, 0.0019, 0.0021, 0.0022], + device='cuda:1'), out_proj_covar=tensor([4.7039e-05, 4.3634e-05, 4.1369e-05, 3.9270e-05, 4.7601e-05, 4.4027e-05, + 4.5833e-05, 5.0386e-05], device='cuda:1') +2023-03-20 19:48:43,316 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=17075.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:48:53,196 INFO [train.py:901] (1/2) Epoch 7, batch 150, loss[loss=0.2217, simple_loss=0.2783, pruned_loss=0.08254, over 7146.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2779, pruned_loss=0.07935, over 764845.72 frames. ], batch size: 41, lr: 2.23e-02, grad_scale: 16.0 +2023-03-20 19:48:59,721 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.829e+02 3.433e+02 4.214e+02 5.230e+02 1.024e+03, threshold=8.429e+02, percent-clipped=1.0 +2023-03-20 19:49:19,122 INFO [train.py:901] (1/2) Epoch 7, batch 200, loss[loss=0.2191, simple_loss=0.2795, pruned_loss=0.07939, over 7274.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2773, pruned_loss=0.07897, over 916944.67 frames. ], batch size: 70, lr: 2.22e-02, grad_scale: 16.0 +2023-03-20 19:49:25,178 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 19:49:28,825 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9014, 3.3699, 3.4076, 3.3049, 3.3581, 3.4163, 3.6484, 3.4278], + device='cuda:1'), covar=tensor([0.0059, 0.0169, 0.0150, 0.0183, 0.0193, 0.0109, 0.0161, 0.0134], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0043, 0.0050, 0.0041, 0.0059, 0.0053, 0.0049, 0.0044], + device='cuda:1'), out_proj_covar=tensor([9.4894e-05, 1.1451e-04, 1.2434e-04, 1.0479e-04, 1.5376e-04, 1.4053e-04, + 1.4088e-04, 1.0816e-04], device='cuda:1') +2023-03-20 19:49:29,206 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 19:49:35,703 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 19:49:36,303 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0922, 1.7330, 1.8562, 1.8456, 1.9587, 1.9474, 2.2145, 1.9814], + device='cuda:1'), covar=tensor([0.0309, 0.0549, 0.0625, 0.0699, 0.0595, 0.0258, 0.0401, 0.0519], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0033, 0.0033, 0.0035, 0.0032, 0.0033, 0.0031, 0.0034], + device='cuda:1'), out_proj_covar=tensor([9.4793e-05, 8.4891e-05, 8.4293e-05, 8.6596e-05, 8.3508e-05, 8.3514e-05, + 8.0357e-05, 8.6198e-05], device='cuda:1') +2023-03-20 19:49:41,650 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17189.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:49:44,512 INFO [train.py:901] (1/2) Epoch 7, batch 250, loss[loss=0.1856, simple_loss=0.2389, pruned_loss=0.06615, over 7186.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2774, pruned_loss=0.07898, over 1031314.61 frames. ], batch size: 39, lr: 2.22e-02, grad_scale: 16.0 +2023-03-20 19:49:49,466 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 19:49:51,488 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.722e+02 3.314e+02 4.064e+02 4.801e+02 1.041e+03, threshold=8.128e+02, percent-clipped=1.0 +2023-03-20 19:49:59,987 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17224.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 19:50:08,329 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 19:50:10,272 INFO [train.py:901] (1/2) Epoch 7, batch 300, loss[loss=0.2153, simple_loss=0.2767, pruned_loss=0.07694, over 7310.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.278, pruned_loss=0.07929, over 1123337.44 frames. ], batch size: 80, lr: 2.22e-02, grad_scale: 16.0 +2023-03-20 19:50:13,033 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17250.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:50:16,902 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 19:50:19,593 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17262.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 19:50:36,769 INFO [train.py:901] (1/2) Epoch 7, batch 350, loss[loss=0.1702, simple_loss=0.2251, pruned_loss=0.05763, over 6985.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2752, pruned_loss=0.07769, over 1192942.87 frames. ], batch size: 35, lr: 2.22e-02, grad_scale: 16.0 +2023-03-20 19:50:42,824 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.792e+02 3.275e+02 4.052e+02 4.742e+02 1.008e+03, threshold=8.103e+02, percent-clipped=2.0 +2023-03-20 19:50:52,932 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 19:51:01,929 INFO [train.py:901] (1/2) Epoch 7, batch 400, loss[loss=0.1979, simple_loss=0.2636, pruned_loss=0.06614, over 7322.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2738, pruned_loss=0.07709, over 1247530.00 frames. ], batch size: 75, lr: 2.21e-02, grad_scale: 16.0 +2023-03-20 19:51:11,172 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17362.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:51:22,193 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17383.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 19:51:28,085 INFO [train.py:901] (1/2) Epoch 7, batch 450, loss[loss=0.2121, simple_loss=0.2773, pruned_loss=0.07345, over 7263.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2748, pruned_loss=0.07762, over 1291544.67 frames. ], batch size: 89, lr: 2.21e-02, grad_scale: 16.0 +2023-03-20 19:51:34,465 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 3.337e+02 4.059e+02 5.111e+02 8.541e+02, threshold=8.117e+02, percent-clipped=3.0 +2023-03-20 19:51:34,509 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 19:51:34,979 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 19:51:42,647 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17423.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:51:53,702 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17444.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 19:51:54,033 INFO [train.py:901] (1/2) Epoch 7, batch 500, loss[loss=0.2042, simple_loss=0.2703, pruned_loss=0.06906, over 7307.00 frames. ], tot_loss[loss=0.214, simple_loss=0.274, pruned_loss=0.07705, over 1323102.84 frames. ], batch size: 80, lr: 2.21e-02, grad_scale: 16.0 +2023-03-20 19:52:07,006 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 19:52:08,563 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 19:52:09,042 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 19:52:11,451 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 19:52:15,829 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 19:52:19,301 INFO [train.py:901] (1/2) Epoch 7, batch 550, loss[loss=0.2369, simple_loss=0.2931, pruned_loss=0.0904, over 7299.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2754, pruned_loss=0.07786, over 1347834.54 frames. ], batch size: 68, lr: 2.20e-02, grad_scale: 16.0 +2023-03-20 19:52:25,332 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.616e+02 3.083e+02 3.892e+02 4.518e+02 9.581e+02, threshold=7.783e+02, percent-clipped=1.0 +2023-03-20 19:52:25,835 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 19:52:31,460 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6400, 3.9218, 4.2688, 4.1727, 4.1401, 4.2195, 4.5670, 4.1296], + device='cuda:1'), covar=tensor([0.0127, 0.0158, 0.0129, 0.0113, 0.0184, 0.0077, 0.0118, 0.0107], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0043, 0.0051, 0.0040, 0.0059, 0.0051, 0.0047, 0.0043], + device='cuda:1'), out_proj_covar=tensor([9.3954e-05, 1.1546e-04, 1.2601e-04, 1.0288e-04, 1.5308e-04, 1.3527e-04, + 1.3263e-04, 1.0652e-04], device='cuda:1') +2023-03-20 19:52:34,462 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17524.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 19:52:34,855 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 19:52:38,343 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 19:52:45,413 INFO [train.py:901] (1/2) Epoch 7, batch 600, loss[loss=0.2299, simple_loss=0.2875, pruned_loss=0.08616, over 7354.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2752, pruned_loss=0.07776, over 1368216.94 frames. ], batch size: 73, lr: 2.20e-02, grad_scale: 8.0 +2023-03-20 19:52:45,419 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 19:52:45,501 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17545.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:52:50,295 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-20 19:52:51,730 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6009, 2.7615, 2.6621, 2.5620, 2.8091, 2.4333, 2.1646, 2.5575], + device='cuda:1'), covar=tensor([0.1953, 0.0291, 0.2246, 0.3740, 0.0833, 0.1130, 0.2971, 0.1514], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0025, 0.0029, 0.0031, 0.0024, 0.0027, 0.0036, 0.0027], + device='cuda:1'), out_proj_covar=tensor([8.6179e-05, 7.5149e-05, 8.7141e-05, 9.1540e-05, 7.8153e-05, 8.6560e-05, + 1.0341e-04, 8.6740e-05], device='cuda:1') +2023-03-20 19:52:53,042 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 +2023-03-20 19:52:54,220 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17562.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 19:52:59,183 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=17572.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 19:53:02,039 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 19:53:09,981 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 19:53:10,476 INFO [train.py:901] (1/2) Epoch 7, batch 650, loss[loss=0.2069, simple_loss=0.2695, pruned_loss=0.07216, over 7369.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2756, pruned_loss=0.07768, over 1386308.24 frames. ], batch size: 73, lr: 2.20e-02, grad_scale: 8.0 +2023-03-20 19:53:17,889 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.063e+02 3.125e+02 3.787e+02 4.523e+02 1.116e+03, threshold=7.575e+02, percent-clipped=2.0 +2023-03-20 19:53:18,972 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=17610.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:53:26,863 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 19:53:37,124 INFO [train.py:901] (1/2) Epoch 7, batch 700, loss[loss=0.2009, simple_loss=0.2674, pruned_loss=0.06719, over 7269.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2753, pruned_loss=0.0774, over 1399227.16 frames. ], batch size: 66, lr: 2.20e-02, grad_scale: 8.0 +2023-03-20 19:53:37,126 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 19:54:02,014 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 19:54:02,498 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 19:54:02,974 INFO [train.py:901] (1/2) Epoch 7, batch 750, loss[loss=0.237, simple_loss=0.2901, pruned_loss=0.09198, over 7259.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2753, pruned_loss=0.07736, over 1410545.35 frames. ], batch size: 52, lr: 2.19e-02, grad_scale: 8.0 +2023-03-20 19:54:05,516 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17700.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:54:09,383 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 3.256e+02 3.901e+02 5.020e+02 1.278e+03, threshold=7.802e+02, percent-clipped=1.0 +2023-03-20 19:54:15,005 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17718.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:54:16,934 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 19:54:17,035 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7887, 3.9297, 3.8063, 3.8905, 3.7415, 4.0105, 3.9197, 3.5898], + device='cuda:1'), covar=tensor([0.0062, 0.0071, 0.0055, 0.0045, 0.0063, 0.0041, 0.0062, 0.0072], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0033, 0.0033, 0.0030, 0.0030, 0.0031, 0.0039, 0.0037], + device='cuda:1'), out_proj_covar=tensor([8.5615e-05, 1.0491e-04, 1.1361e-04, 8.9700e-05, 9.3743e-05, 9.9777e-05, + 1.3149e-04, 1.1621e-04], device='cuda:1') +2023-03-20 19:54:17,193 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 +2023-03-20 19:54:18,382 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.12 vs. limit=5.0 +2023-03-20 19:54:20,960 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 19:54:21,051 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17730.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:54:25,528 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17739.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 19:54:26,049 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7193, 4.4324, 4.2202, 3.9410, 4.2665, 2.9129, 2.0564, 4.6601], + device='cuda:1'), covar=tensor([0.0011, 0.0060, 0.0060, 0.0050, 0.0025, 0.0349, 0.0582, 0.0045], + device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0051, 0.0071, 0.0054, 0.0061, 0.0090, 0.0098, 0.0058], + device='cuda:1'), out_proj_covar=tensor([6.8201e-05, 8.0252e-05, 1.0257e-04, 8.1175e-05, 8.2286e-05, 1.3202e-04, + 1.4429e-04, 8.2810e-05], device='cuda:1') +2023-03-20 19:54:27,488 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 19:54:28,403 INFO [train.py:901] (1/2) Epoch 7, batch 800, loss[loss=0.2625, simple_loss=0.3154, pruned_loss=0.1048, over 7253.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2752, pruned_loss=0.07736, over 1417156.73 frames. ], batch size: 89, lr: 2.19e-02, grad_scale: 8.0 +2023-03-20 19:54:28,441 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 19:54:36,480 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17761.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:54:38,294 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 19:54:51,906 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17791.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:54:53,738 INFO [train.py:901] (1/2) Epoch 7, batch 850, loss[loss=0.246, simple_loss=0.3084, pruned_loss=0.09176, over 7250.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2755, pruned_loss=0.07714, over 1424931.81 frames. ], batch size: 89, lr: 2.19e-02, grad_scale: 8.0 +2023-03-20 19:54:57,080 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 19:54:57,089 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 19:55:01,228 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.727e+02 3.034e+02 3.726e+02 4.781e+02 1.153e+03, threshold=7.452e+02, percent-clipped=1.0 +2023-03-20 19:55:03,233 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 19:55:06,727 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 19:55:19,515 INFO [train.py:901] (1/2) Epoch 7, batch 900, loss[loss=0.2141, simple_loss=0.273, pruned_loss=0.07763, over 7331.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2754, pruned_loss=0.07721, over 1429793.53 frames. ], batch size: 75, lr: 2.18e-02, grad_scale: 8.0 +2023-03-20 19:55:19,613 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17845.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:55:24,641 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.7626, 0.6955, 0.8798, 1.0298, 0.9069, 0.8807, 0.9229, 0.8158], + device='cuda:1'), covar=tensor([0.0980, 0.3446, 0.0647, 0.0437, 0.1124, 0.1608, 0.0748, 0.2088], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0035, 0.0031, 0.0030, 0.0033, 0.0029, 0.0034, 0.0033], + device='cuda:1'), out_proj_covar=tensor([5.3535e-05, 7.3077e-05, 5.1097e-05, 4.9961e-05, 5.9673e-05, 5.5541e-05, + 6.1443e-05, 6.1278e-05], device='cuda:1') +2023-03-20 19:55:44,955 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 19:55:44,987 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=17893.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:55:45,954 INFO [train.py:901] (1/2) Epoch 7, batch 950, loss[loss=0.148, simple_loss=0.1961, pruned_loss=0.04996, over 5961.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2746, pruned_loss=0.07683, over 1430104.39 frames. ], batch size: 26, lr: 2.18e-02, grad_scale: 8.0 +2023-03-20 19:55:52,440 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.338e+02 3.363e+02 3.920e+02 4.922e+02 1.038e+03, threshold=7.839e+02, percent-clipped=2.0 +2023-03-20 19:55:56,825 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.10 vs. limit=2.0 +2023-03-20 19:56:08,909 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 19:56:10,870 INFO [train.py:901] (1/2) Epoch 7, batch 1000, loss[loss=0.2179, simple_loss=0.2799, pruned_loss=0.07799, over 7287.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2747, pruned_loss=0.07681, over 1434153.26 frames. ], batch size: 70, lr: 2.18e-02, grad_scale: 8.0 +2023-03-20 19:56:19,232 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17960.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:56:19,901 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-03-20 19:56:28,613 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 19:56:35,870 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 +2023-03-20 19:56:37,045 INFO [train.py:901] (1/2) Epoch 7, batch 1050, loss[loss=0.2115, simple_loss=0.2677, pruned_loss=0.07771, over 7213.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2753, pruned_loss=0.07699, over 1433957.52 frames. ], batch size: 45, lr: 2.18e-02, grad_scale: 8.0 +2023-03-20 19:56:43,870 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 3.096e+02 3.604e+02 4.422e+02 1.016e+03, threshold=7.208e+02, percent-clipped=3.0 +2023-03-20 19:56:49,038 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18018.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:56:50,581 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18021.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:56:51,982 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 19:56:55,507 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 19:56:59,633 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18039.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 19:57:03,124 INFO [train.py:901] (1/2) Epoch 7, batch 1100, loss[loss=0.2056, simple_loss=0.2729, pruned_loss=0.06912, over 7283.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2743, pruned_loss=0.07657, over 1434386.37 frames. ], batch size: 70, lr: 2.17e-02, grad_scale: 8.0 +2023-03-20 19:57:08,649 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18056.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:57:13,601 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=18066.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:57:23,810 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5899, 4.5592, 4.4836, 4.8344, 4.9381, 4.9139, 4.4224, 4.4743], + device='cuda:1'), covar=tensor([0.0750, 0.1686, 0.1874, 0.0970, 0.0494, 0.0896, 0.0476, 0.0605], + device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0208, 0.0190, 0.0174, 0.0140, 0.0220, 0.0127, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 19:57:24,274 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 19:57:24,331 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18086.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:57:24,765 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 19:57:24,832 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=18087.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 19:57:27,429 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-20 19:57:27,478 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-20 19:57:28,653 INFO [train.py:901] (1/2) Epoch 7, batch 1150, loss[loss=0.2065, simple_loss=0.2766, pruned_loss=0.06822, over 7355.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2743, pruned_loss=0.0766, over 1433258.60 frames. ], batch size: 63, lr: 2.17e-02, grad_scale: 8.0 +2023-03-20 19:57:35,151 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.225e+02 2.927e+02 3.944e+02 4.726e+02 8.317e+02, threshold=7.887e+02, percent-clipped=4.0 +2023-03-20 19:57:36,639 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 19:57:37,176 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 19:57:37,759 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7552, 3.3350, 3.5529, 3.4392, 3.3758, 3.4667, 3.6381, 3.3881], + device='cuda:1'), covar=tensor([0.0125, 0.0166, 0.0145, 0.0162, 0.0181, 0.0107, 0.0163, 0.0126], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0045, 0.0054, 0.0042, 0.0062, 0.0055, 0.0050, 0.0045], + device='cuda:1'), out_proj_covar=tensor([9.8497e-05, 1.1945e-04, 1.3530e-04, 1.0692e-04, 1.6221e-04, 1.4946e-04, + 1.4237e-04, 1.1126e-04], device='cuda:1') +2023-03-20 19:57:54,426 INFO [train.py:901] (1/2) Epoch 7, batch 1200, loss[loss=0.2152, simple_loss=0.2855, pruned_loss=0.07241, over 7279.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2739, pruned_loss=0.07595, over 1436227.74 frames. ], batch size: 66, lr: 2.17e-02, grad_scale: 8.0 +2023-03-20 19:58:09,723 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18174.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:58:10,269 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18175.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:58:11,112 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 19:58:14,904 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-03-20 19:58:20,173 INFO [train.py:901] (1/2) Epoch 7, batch 1250, loss[loss=0.1753, simple_loss=0.219, pruned_loss=0.06578, over 6254.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2737, pruned_loss=0.07618, over 1437487.83 frames. ], batch size: 26, lr: 2.16e-02, grad_scale: 8.0 +2023-03-20 19:58:23,255 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 +2023-03-20 19:58:26,361 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-03-20 19:58:27,026 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.616e+02 3.110e+02 3.917e+02 4.843e+02 1.048e+03, threshold=7.833e+02, percent-clipped=3.0 +2023-03-20 19:58:35,820 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 19:58:37,576 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.11 vs. limit=2.0 +2023-03-20 19:58:40,308 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 19:58:41,313 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 19:58:41,448 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18235.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:58:41,948 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18236.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:58:46,312 INFO [train.py:901] (1/2) Epoch 7, batch 1300, loss[loss=0.1539, simple_loss=0.2179, pruned_loss=0.04501, over 7061.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2738, pruned_loss=0.07606, over 1438459.65 frames. ], batch size: 35, lr: 2.16e-02, grad_scale: 8.0 +2023-03-20 19:58:51,554 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4668, 4.1375, 3.8224, 4.1256, 3.0859, 2.8568, 4.3496, 3.6813], + device='cuda:1'), covar=tensor([0.0069, 0.0071, 0.0090, 0.0045, 0.0218, 0.0306, 0.0071, 0.0240], + device='cuda:1'), in_proj_covar=tensor([0.0218, 0.0204, 0.0223, 0.0176, 0.0279, 0.0286, 0.0216, 0.0289], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 19:59:04,330 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 19:59:06,327 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 19:59:09,706 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 19:59:11,661 INFO [train.py:901] (1/2) Epoch 7, batch 1350, loss[loss=0.2094, simple_loss=0.2783, pruned_loss=0.07022, over 7293.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2747, pruned_loss=0.07672, over 1438364.93 frames. ], batch size: 86, lr: 2.16e-02, grad_scale: 8.0 +2023-03-20 19:59:18,814 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 3.114e+02 3.862e+02 4.620e+02 8.762e+02, threshold=7.723e+02, percent-clipped=2.0 +2023-03-20 19:59:20,348 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 19:59:22,909 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18316.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:59:27,231 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-03-20 19:59:36,703 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 19:59:37,879 INFO [train.py:901] (1/2) Epoch 7, batch 1400, loss[loss=0.2195, simple_loss=0.2865, pruned_loss=0.07621, over 7320.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2739, pruned_loss=0.07611, over 1440629.19 frames. ], batch size: 59, lr: 2.16e-02, grad_scale: 8.0 +2023-03-20 19:59:43,596 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18356.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 19:59:45,598 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1552, 4.2008, 4.0708, 4.6079, 4.5798, 4.5123, 4.1555, 4.1047], + device='cuda:1'), covar=tensor([0.0936, 0.2182, 0.2331, 0.1049, 0.0546, 0.1309, 0.0787, 0.1011], + device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0212, 0.0193, 0.0174, 0.0142, 0.0229, 0.0132, 0.0150], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-03-20 19:59:51,605 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 19:59:58,742 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18386.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:00:03,667 INFO [train.py:901] (1/2) Epoch 7, batch 1450, loss[loss=0.2122, simple_loss=0.279, pruned_loss=0.0727, over 7261.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2747, pruned_loss=0.07635, over 1441120.43 frames. ], batch size: 89, lr: 2.15e-02, grad_scale: 8.0 +2023-03-20 20:00:08,548 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=18404.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:00:10,482 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.072e+02 3.218e+02 3.702e+02 4.494e+02 7.594e+02, threshold=7.404e+02, percent-clipped=0.0 +2023-03-20 20:00:14,167 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3236, 4.2284, 4.1190, 4.7365, 4.6599, 4.6693, 4.1933, 4.2194], + device='cuda:1'), covar=tensor([0.0860, 0.1662, 0.1829, 0.1022, 0.0558, 0.1124, 0.0667, 0.0850], + device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0211, 0.0191, 0.0173, 0.0141, 0.0223, 0.0129, 0.0149], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-03-20 20:00:16,155 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 20:00:24,324 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=18434.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:00:29,765 INFO [train.py:901] (1/2) Epoch 7, batch 1500, loss[loss=0.2208, simple_loss=0.2845, pruned_loss=0.07859, over 7348.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2743, pruned_loss=0.07595, over 1443642.20 frames. ], batch size: 54, lr: 2.15e-02, grad_scale: 8.0 +2023-03-20 20:00:31,841 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 20:00:53,290 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18490.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:00:55,596 INFO [train.py:901] (1/2) Epoch 7, batch 1550, loss[loss=0.2739, simple_loss=0.3187, pruned_loss=0.1146, over 6804.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2733, pruned_loss=0.07552, over 1442853.89 frames. ], batch size: 107, lr: 2.15e-02, grad_scale: 8.0 +2023-03-20 20:00:56,641 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 20:01:02,706 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.878e+02 3.000e+02 3.800e+02 4.612e+02 7.440e+02, threshold=7.601e+02, percent-clipped=1.0 +2023-03-20 20:01:02,858 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18508.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 20:01:09,071 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 +2023-03-20 20:01:13,801 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18530.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:01:14,314 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18531.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:01:21,262 INFO [train.py:901] (1/2) Epoch 7, batch 1600, loss[loss=0.2299, simple_loss=0.295, pruned_loss=0.08243, over 7276.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2728, pruned_loss=0.07557, over 1441722.61 frames. ], batch size: 77, lr: 2.15e-02, grad_scale: 8.0 +2023-03-20 20:01:24,333 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18551.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:01:27,697 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 20:01:28,693 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 20:01:32,304 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 20:01:33,907 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18569.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 20:01:34,536 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 +2023-03-20 20:01:39,813 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5562, 3.0360, 2.6913, 2.8135, 2.7333, 2.5462, 1.9785, 2.6163], + device='cuda:1'), covar=tensor([0.2188, 0.0248, 0.1209, 0.1681, 0.0842, 0.1790, 0.2141, 0.1701], + device='cuda:1'), in_proj_covar=tensor([0.0026, 0.0023, 0.0026, 0.0028, 0.0025, 0.0027, 0.0035, 0.0027], + device='cuda:1'), out_proj_covar=tensor([8.6769e-05, 7.3951e-05, 8.6777e-05, 9.1179e-05, 8.2475e-05, 8.8762e-05, + 1.0459e-04, 8.9168e-05], device='cuda:1') +2023-03-20 20:01:41,137 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 20:01:45,087 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 20:01:46,555 INFO [train.py:901] (1/2) Epoch 7, batch 1650, loss[loss=0.2094, simple_loss=0.2688, pruned_loss=0.07497, over 7311.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2738, pruned_loss=0.0761, over 1442781.90 frames. ], batch size: 49, lr: 2.14e-02, grad_scale: 8.0 +2023-03-20 20:01:54,019 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 3.211e+02 3.996e+02 4.881e+02 1.178e+03, threshold=7.992e+02, percent-clipped=2.0 +2023-03-20 20:01:54,043 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 20:01:58,128 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18616.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:02:10,311 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 20:02:12,337 INFO [train.py:901] (1/2) Epoch 7, batch 1700, loss[loss=0.2105, simple_loss=0.2743, pruned_loss=0.07336, over 7313.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2735, pruned_loss=0.07598, over 1441704.09 frames. ], batch size: 80, lr: 2.14e-02, grad_scale: 8.0 +2023-03-20 20:02:14,062 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2908, 1.1926, 1.2650, 0.9469, 1.2550, 1.0098, 1.3126, 0.8558], + device='cuda:1'), covar=tensor([0.0125, 0.0173, 0.0216, 0.0129, 0.0207, 0.0148, 0.0177, 0.0238], + device='cuda:1'), in_proj_covar=tensor([0.0018, 0.0017, 0.0017, 0.0016, 0.0019, 0.0017, 0.0017, 0.0020], + device='cuda:1'), out_proj_covar=tensor([2.1982e-05, 1.9614e-05, 2.2772e-05, 1.8034e-05, 2.1437e-05, 2.0340e-05, + 2.1109e-05, 2.8890e-05], device='cuda:1') +2023-03-20 20:02:14,424 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 20:02:22,608 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=18664.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:02:25,714 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 20:02:27,794 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18674.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:02:30,355 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2924, 1.8401, 2.0909, 1.9515, 2.0960, 1.7900, 2.3692, 2.0851], + device='cuda:1'), covar=tensor([0.0694, 0.1161, 0.0956, 0.1618, 0.1362, 0.1008, 0.1174, 0.1602], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0033, 0.0035, 0.0034, 0.0032, 0.0034, 0.0031, 0.0034], + device='cuda:1'), out_proj_covar=tensor([1.0143e-04, 8.9793e-05, 9.3330e-05, 8.9725e-05, 8.9098e-05, 9.0041e-05, + 8.5711e-05, 8.9498e-05], device='cuda:1') +2023-03-20 20:02:33,601 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 +2023-03-20 20:02:38,755 INFO [train.py:901] (1/2) Epoch 7, batch 1750, loss[loss=0.1947, simple_loss=0.2544, pruned_loss=0.0675, over 7302.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2724, pruned_loss=0.07513, over 1443197.47 frames. ], batch size: 68, lr: 2.14e-02, grad_scale: 8.0 +2023-03-20 20:02:45,510 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.854e+02 2.856e+02 3.595e+02 4.544e+02 1.273e+03, threshold=7.190e+02, percent-clipped=1.0 +2023-03-20 20:02:50,994 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 20:02:52,002 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 20:02:59,066 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18735.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:03:01,568 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5977, 5.0200, 4.9530, 4.9382, 4.7369, 4.6165, 5.0281, 4.8189], + device='cuda:1'), covar=tensor([0.0349, 0.0383, 0.0475, 0.0473, 0.0426, 0.0283, 0.0338, 0.0575], + device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0142, 0.0116, 0.0104, 0.0094, 0.0130, 0.0120, 0.0097], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 20:03:04,589 INFO [train.py:901] (1/2) Epoch 7, batch 1800, loss[loss=0.2609, simple_loss=0.3158, pruned_loss=0.1029, over 7258.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2732, pruned_loss=0.07547, over 1444027.11 frames. ], batch size: 64, lr: 2.13e-02, grad_scale: 8.0 +2023-03-20 20:03:15,018 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 20:03:18,158 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2549, 1.6176, 2.0446, 1.1218, 1.2609, 1.4324, 1.0574, 1.1892], + device='cuda:1'), covar=tensor([0.0173, 0.0177, 0.0087, 0.0056, 0.0186, 0.0223, 0.0288, 0.0309], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0018, 0.0018, 0.0019, 0.0018, 0.0017, 0.0019, 0.0019], + device='cuda:1'), out_proj_covar=tensor([4.5025e-05, 4.0396e-05, 3.9396e-05, 3.6031e-05, 4.2452e-05, 4.0352e-05, + 4.2545e-05, 4.5912e-05], device='cuda:1') +2023-03-20 20:03:29,219 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 20:03:30,209 INFO [train.py:901] (1/2) Epoch 7, batch 1850, loss[loss=0.1975, simple_loss=0.2667, pruned_loss=0.06418, over 7343.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.273, pruned_loss=0.07528, over 1444301.13 frames. ], batch size: 61, lr: 2.13e-02, grad_scale: 8.0 +2023-03-20 20:03:37,097 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.910e+02 3.015e+02 3.933e+02 4.902e+02 8.513e+02, threshold=7.867e+02, percent-clipped=4.0 +2023-03-20 20:03:39,182 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 20:03:48,846 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18830.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:03:49,308 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18831.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:03:49,349 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3321, 1.5616, 2.2527, 1.5128, 1.3281, 1.0586, 1.1586, 1.3626], + device='cuda:1'), covar=tensor([0.0259, 0.0325, 0.0264, 0.0066, 0.0370, 0.0388, 0.0266, 0.0232], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0018, 0.0018, 0.0019, 0.0018, 0.0017, 0.0019, 0.0019], + device='cuda:1'), out_proj_covar=tensor([4.5786e-05, 4.0858e-05, 3.9679e-05, 3.5945e-05, 4.3601e-05, 4.0470e-05, + 4.2187e-05, 4.6011e-05], device='cuda:1') +2023-03-20 20:03:53,981 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-03-20 20:03:55,669 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 20:03:56,136 INFO [train.py:901] (1/2) Epoch 7, batch 1900, loss[loss=0.2106, simple_loss=0.2666, pruned_loss=0.07731, over 7280.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2734, pruned_loss=0.07558, over 1444806.63 frames. ], batch size: 47, lr: 2.13e-02, grad_scale: 8.0 +2023-03-20 20:03:56,739 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18846.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:04:06,311 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18864.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 20:04:10,918 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0558, 0.8720, 1.0943, 1.1488, 1.1297, 1.0912, 1.0236, 1.1901], + device='cuda:1'), covar=tensor([0.0996, 0.1688, 0.0694, 0.0964, 0.1615, 0.0689, 0.0726, 0.1839], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0035, 0.0032, 0.0030, 0.0030, 0.0030, 0.0035, 0.0032], + device='cuda:1'), out_proj_covar=tensor([5.2996e-05, 7.1906e-05, 5.4955e-05, 5.3270e-05, 5.8187e-05, 5.7631e-05, + 6.4596e-05, 6.1295e-05], device='cuda:1') +2023-03-20 20:04:13,351 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=18878.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:04:13,826 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=18879.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:04:16,154 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 +2023-03-20 20:04:18,409 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7623, 3.9609, 3.6899, 3.8113, 3.5871, 3.9019, 4.3451, 4.3103], + device='cuda:1'), covar=tensor([0.0238, 0.0171, 0.0208, 0.0214, 0.0448, 0.0246, 0.0196, 0.0175], + device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0092, 0.0085, 0.0096, 0.0094, 0.0075, 0.0074, 0.0074], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 20:04:19,827 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 20:04:21,787 INFO [train.py:901] (1/2) Epoch 7, batch 1950, loss[loss=0.1902, simple_loss=0.2597, pruned_loss=0.06034, over 7298.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2733, pruned_loss=0.0754, over 1443722.26 frames. ], batch size: 80, lr: 2.13e-02, grad_scale: 8.0 +2023-03-20 20:04:26,097 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-03-20 20:04:28,237 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.263e+02 3.431e+02 3.888e+02 4.453e+02 9.116e+02, threshold=7.775e+02, percent-clipped=1.0 +2023-03-20 20:04:30,809 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 20:04:35,432 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 20:04:35,936 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 20:04:40,025 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18930.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:04:47,366 INFO [train.py:901] (1/2) Epoch 7, batch 2000, loss[loss=0.2165, simple_loss=0.2826, pruned_loss=0.07518, over 7207.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.273, pruned_loss=0.07558, over 1440290.89 frames. ], batch size: 93, lr: 2.12e-02, grad_scale: 8.0 +2023-03-20 20:04:53,099 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 20:05:03,031 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 +2023-03-20 20:05:04,206 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 20:05:11,440 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18991.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:05:12,783 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 20:05:13,298 INFO [train.py:901] (1/2) Epoch 7, batch 2050, loss[loss=0.1853, simple_loss=0.2469, pruned_loss=0.06186, over 7257.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2723, pruned_loss=0.07491, over 1440830.31 frames. ], batch size: 47, lr: 2.12e-02, grad_scale: 8.0 +2023-03-20 20:05:16,254 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3911, 4.2950, 4.1518, 4.5907, 4.6317, 4.6233, 4.0184, 4.1137], + device='cuda:1'), covar=tensor([0.0931, 0.1898, 0.2167, 0.1278, 0.0735, 0.1367, 0.0704, 0.0958], + device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0223, 0.0201, 0.0189, 0.0144, 0.0237, 0.0133, 0.0158], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:1') +2023-03-20 20:05:20,082 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 3.151e+02 3.634e+02 4.204e+02 9.564e+02, threshold=7.267e+02, percent-clipped=4.0 +2023-03-20 20:05:31,880 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19030.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:05:39,804 INFO [train.py:901] (1/2) Epoch 7, batch 2100, loss[loss=0.1899, simple_loss=0.2535, pruned_loss=0.0631, over 7233.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2725, pruned_loss=0.07489, over 1443009.87 frames. ], batch size: 45, lr: 2.12e-02, grad_scale: 8.0 +2023-03-20 20:05:46,880 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 20:05:49,909 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 20:05:50,326 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 +2023-03-20 20:06:05,010 INFO [train.py:901] (1/2) Epoch 7, batch 2150, loss[loss=0.1786, simple_loss=0.2467, pruned_loss=0.05527, over 7243.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2722, pruned_loss=0.07438, over 1444654.50 frames. ], batch size: 55, lr: 2.12e-02, grad_scale: 8.0 +2023-03-20 20:06:05,817 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8321, 3.1117, 2.4737, 3.9566, 1.8987, 3.2315, 1.5582, 3.8371], + device='cuda:1'), covar=tensor([0.0045, 0.0433, 0.1613, 0.0033, 0.4331, 0.0054, 0.1066, 0.0055], + device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0248, 0.0321, 0.0122, 0.0303, 0.0139, 0.0264, 0.0154], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 20:06:08,040 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.00 vs. limit=2.0 +2023-03-20 20:06:10,998 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.09 vs. limit=2.0 +2023-03-20 20:06:12,197 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.027e+02 3.142e+02 3.714e+02 4.816e+02 1.137e+03, threshold=7.428e+02, percent-clipped=3.0 +2023-03-20 20:06:23,667 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19129.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 20:06:31,457 INFO [train.py:901] (1/2) Epoch 7, batch 2200, loss[loss=0.1737, simple_loss=0.2327, pruned_loss=0.05741, over 7173.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2721, pruned_loss=0.07421, over 1442916.25 frames. ], batch size: 39, lr: 2.11e-02, grad_scale: 8.0 +2023-03-20 20:06:32,096 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19146.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:06:32,878 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.10 vs. limit=5.0 +2023-03-20 20:06:36,566 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 20:06:36,725 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1116, 2.2690, 2.1297, 3.1030, 2.8709, 3.1406, 2.9718, 3.0152], + device='cuda:1'), covar=tensor([0.1004, 0.0427, 0.1250, 0.0235, 0.0020, 0.0055, 0.0025, 0.0030], + device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0205, 0.0249, 0.0188, 0.0106, 0.0105, 0.0110, 0.0107], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 20:06:40,130 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6587, 5.0519, 5.1497, 4.9629, 4.8349, 4.5945, 5.0690, 4.8717], + device='cuda:1'), covar=tensor([0.0324, 0.0339, 0.0256, 0.0375, 0.0333, 0.0254, 0.0317, 0.0472], + device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0138, 0.0111, 0.0103, 0.0091, 0.0127, 0.0116, 0.0091], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 20:06:41,211 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19164.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 20:06:54,939 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19190.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 20:06:56,873 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=19194.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:06:57,303 INFO [train.py:901] (1/2) Epoch 7, batch 2250, loss[loss=0.2662, simple_loss=0.3175, pruned_loss=0.1075, over 6727.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2714, pruned_loss=0.07402, over 1442738.82 frames. ], batch size: 106, lr: 2.11e-02, grad_scale: 8.0 +2023-03-20 20:07:04,212 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.690e+02 2.891e+02 3.634e+02 4.674e+02 1.266e+03, threshold=7.268e+02, percent-clipped=2.0 +2023-03-20 20:07:06,338 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=19212.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 20:07:10,992 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 20:07:11,516 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 20:07:14,256 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 +2023-03-20 20:07:23,531 INFO [train.py:901] (1/2) Epoch 7, batch 2300, loss[loss=0.2072, simple_loss=0.2659, pruned_loss=0.0743, over 7307.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2722, pruned_loss=0.07414, over 1443258.19 frames. ], batch size: 83, lr: 2.11e-02, grad_scale: 8.0 +2023-03-20 20:07:23,534 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 20:07:26,188 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5479, 3.6869, 3.3674, 3.4678, 3.4704, 3.4732, 3.8258, 3.7662], + device='cuda:1'), covar=tensor([0.0190, 0.0156, 0.0217, 0.0209, 0.0278, 0.0243, 0.0236, 0.0215], + device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0091, 0.0083, 0.0095, 0.0092, 0.0071, 0.0074, 0.0073], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 20:07:27,768 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5086, 3.6989, 2.5207, 3.9139, 3.2971, 3.6922, 2.1796, 2.1467], + device='cuda:1'), covar=tensor([0.0066, 0.0197, 0.0655, 0.0125, 0.0119, 0.0169, 0.1029, 0.0607], + device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0168, 0.0271, 0.0151, 0.0164, 0.0155, 0.0256, 0.0259], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 20:07:44,941 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19286.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:07:48,481 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6804, 3.8038, 3.3955, 3.5743, 3.5399, 3.5132, 3.9130, 3.9052], + device='cuda:1'), covar=tensor([0.0180, 0.0166, 0.0276, 0.0214, 0.0306, 0.0336, 0.0264, 0.0228], + device='cuda:1'), in_proj_covar=tensor([0.0089, 0.0091, 0.0084, 0.0096, 0.0094, 0.0072, 0.0074, 0.0073], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 20:07:49,334 INFO [train.py:901] (1/2) Epoch 7, batch 2350, loss[loss=0.2253, simple_loss=0.2872, pruned_loss=0.08164, over 7340.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2722, pruned_loss=0.07433, over 1441919.59 frames. ], batch size: 54, lr: 2.11e-02, grad_scale: 8.0 +2023-03-20 20:07:56,476 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.976e+02 3.198e+02 3.776e+02 4.552e+02 1.396e+03, threshold=7.552e+02, percent-clipped=7.0 +2023-03-20 20:08:02,408 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-03-20 20:08:07,648 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19330.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:08:09,630 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 20:08:15,099 INFO [train.py:901] (1/2) Epoch 7, batch 2400, loss[loss=0.212, simple_loss=0.2851, pruned_loss=0.06944, over 7285.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2719, pruned_loss=0.0739, over 1443562.93 frames. ], batch size: 66, lr: 2.10e-02, grad_scale: 8.0 +2023-03-20 20:08:15,628 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 20:08:20,979 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 +2023-03-20 20:08:26,348 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 20:08:28,841 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 20:08:32,472 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=19378.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:08:39,227 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0097, 2.0832, 1.9892, 3.2637, 2.5335, 2.9794, 3.0850, 2.7962], + device='cuda:1'), covar=tensor([0.1127, 0.0602, 0.1497, 0.0221, 0.0023, 0.0039, 0.0042, 0.0031], + device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0216, 0.0268, 0.0197, 0.0112, 0.0112, 0.0115, 0.0112], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 20:08:41,189 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19394.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:08:41,563 INFO [train.py:901] (1/2) Epoch 7, batch 2450, loss[loss=0.2229, simple_loss=0.2807, pruned_loss=0.08259, over 7216.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2715, pruned_loss=0.07368, over 1442868.95 frames. ], batch size: 50, lr: 2.10e-02, grad_scale: 8.0 +2023-03-20 20:08:47,243 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 +2023-03-20 20:08:48,355 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.646e+02 3.092e+02 3.904e+02 5.234e+02 1.110e+03, threshold=7.808e+02, percent-clipped=4.0 +2023-03-20 20:08:54,842 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 20:09:07,596 INFO [train.py:901] (1/2) Epoch 7, batch 2500, loss[loss=0.2057, simple_loss=0.2759, pruned_loss=0.06781, over 7337.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2713, pruned_loss=0.07355, over 1444135.41 frames. ], batch size: 61, lr: 2.10e-02, grad_scale: 8.0 +2023-03-20 20:09:12,784 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19455.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:09:21,266 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 20:09:21,933 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0294, 0.7411, 1.1050, 1.3302, 1.3181, 1.2655, 1.1911, 1.1342], + device='cuda:1'), covar=tensor([0.0800, 0.1204, 0.0368, 0.0483, 0.0885, 0.0614, 0.0429, 0.1526], + device='cuda:1'), in_proj_covar=tensor([0.0026, 0.0032, 0.0026, 0.0028, 0.0028, 0.0027, 0.0032, 0.0032], + device='cuda:1'), out_proj_covar=tensor([5.0819e-05, 6.8272e-05, 4.7009e-05, 4.9874e-05, 5.5567e-05, 5.3690e-05, + 6.0160e-05, 6.0817e-05], device='cuda:1') +2023-03-20 20:09:28,486 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19485.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 20:09:33,346 INFO [train.py:901] (1/2) Epoch 7, batch 2550, loss[loss=0.1985, simple_loss=0.256, pruned_loss=0.07047, over 6991.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2713, pruned_loss=0.07374, over 1442026.68 frames. ], batch size: 35, lr: 2.10e-02, grad_scale: 8.0 +2023-03-20 20:09:39,864 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.969e+02 3.253e+02 3.755e+02 4.518e+02 7.754e+02, threshold=7.511e+02, percent-clipped=0.0 +2023-03-20 20:09:59,157 INFO [train.py:901] (1/2) Epoch 7, batch 2600, loss[loss=0.2196, simple_loss=0.2795, pruned_loss=0.07985, over 7280.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2708, pruned_loss=0.07361, over 1440486.54 frames. ], batch size: 77, lr: 2.09e-02, grad_scale: 16.0 +2023-03-20 20:10:19,586 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19586.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:10:23,820 INFO [train.py:901] (1/2) Epoch 7, batch 2650, loss[loss=0.2114, simple_loss=0.2773, pruned_loss=0.07277, over 7267.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2711, pruned_loss=0.07361, over 1443006.09 frames. ], batch size: 64, lr: 2.09e-02, grad_scale: 16.0 +2023-03-20 20:10:31,140 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.231e+02 3.103e+02 3.938e+02 4.730e+02 1.209e+03, threshold=7.875e+02, percent-clipped=4.0 +2023-03-20 20:10:36,210 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19618.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:10:39,257 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1222, 1.2959, 1.5888, 0.8797, 1.0145, 0.9819, 1.1421, 0.9583], + device='cuda:1'), covar=tensor([0.0273, 0.0214, 0.0246, 0.0092, 0.0452, 0.0470, 0.0098, 0.0145], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0018, 0.0019, 0.0019, 0.0020, 0.0018, 0.0019, 0.0019], + device='cuda:1'), out_proj_covar=tensor([4.5546e-05, 4.2060e-05, 4.1173e-05, 3.7055e-05, 4.5904e-05, 4.1410e-05, + 4.2217e-05, 4.6377e-05], device='cuda:1') +2023-03-20 20:10:44,155 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=19634.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:10:48,614 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2185, 1.1592, 1.6053, 0.8680, 1.1537, 1.1154, 1.0954, 1.0635], + device='cuda:1'), covar=tensor([0.0189, 0.0209, 0.0140, 0.0074, 0.0205, 0.0237, 0.0136, 0.0131], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0018, 0.0018, 0.0019, 0.0019, 0.0018, 0.0019, 0.0019], + device='cuda:1'), out_proj_covar=tensor([4.4476e-05, 4.1522e-05, 4.0371e-05, 3.6547e-05, 4.4842e-05, 4.0896e-05, + 4.1724e-05, 4.5663e-05], device='cuda:1') +2023-03-20 20:10:49,452 INFO [train.py:901] (1/2) Epoch 7, batch 2700, loss[loss=0.2315, simple_loss=0.2945, pruned_loss=0.0842, over 7247.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2704, pruned_loss=0.07308, over 1442178.86 frames. ], batch size: 55, lr: 2.09e-02, grad_scale: 16.0 +2023-03-20 20:11:06,095 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19679.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:11:13,899 INFO [train.py:901] (1/2) Epoch 7, batch 2750, loss[loss=0.1953, simple_loss=0.2643, pruned_loss=0.06319, over 7349.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2708, pruned_loss=0.07325, over 1443581.15 frames. ], batch size: 61, lr: 2.09e-02, grad_scale: 16.0 +2023-03-20 20:11:15,989 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5616, 3.2334, 3.4055, 3.2327, 3.3382, 3.4008, 3.5490, 3.3341], + device='cuda:1'), covar=tensor([0.0100, 0.0162, 0.0120, 0.0156, 0.0176, 0.0093, 0.0150, 0.0112], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0047, 0.0054, 0.0042, 0.0064, 0.0056, 0.0050, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 20:11:20,357 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.753e+02 2.910e+02 3.544e+02 4.508e+02 1.425e+03, threshold=7.087e+02, percent-clipped=5.0 +2023-03-20 20:11:38,344 INFO [train.py:901] (1/2) Epoch 7, batch 2800, loss[loss=0.1878, simple_loss=0.243, pruned_loss=0.06635, over 6949.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.271, pruned_loss=0.07344, over 1444365.05 frames. ], batch size: 35, lr: 2.08e-02, grad_scale: 16.0 +2023-03-20 20:11:39,463 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2164, 2.3311, 1.9920, 3.5510, 1.6165, 2.9202, 1.6977, 3.3864], + device='cuda:1'), covar=tensor([0.0053, 0.0758, 0.1654, 0.0038, 0.4196, 0.0061, 0.0993, 0.0065], + device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0250, 0.0322, 0.0125, 0.0301, 0.0136, 0.0264, 0.0158], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 20:11:40,812 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19750.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:11:49,051 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-03-20 20:12:03,572 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 20:12:03,645 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 20:12:03,672 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 20:12:03,689 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 20:12:03,864 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 20:12:03,879 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 20:12:03,890 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 20:12:03,991 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 20:12:04,065 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 20:12:04,079 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 20:12:04,097 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 20:12:04,263 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 20:12:04,268 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 20:12:04,630 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 20:12:04,774 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 20:12:04,777 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 20:12:04,834 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 20:12:12,424 INFO [train.py:901] (1/2) Epoch 8, batch 0, loss[loss=0.1925, simple_loss=0.2551, pruned_loss=0.06495, over 7308.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2551, pruned_loss=0.06495, over 7308.00 frames. ], batch size: 86, lr: 2.00e-02, grad_scale: 16.0 +2023-03-20 20:12:12,424 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 20:12:34,444 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7156, 4.1972, 4.0026, 4.4174, 4.5135, 4.4421, 4.1157, 4.0251], + device='cuda:1'), covar=tensor([0.1051, 0.1761, 0.2137, 0.1451, 0.0604, 0.1545, 0.0769, 0.0917], + device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0231, 0.0203, 0.0186, 0.0153, 0.0243, 0.0132, 0.0160], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002], + device='cuda:1') +2023-03-20 20:12:38,315 INFO [train.py:935] (1/2) Epoch 8, validation: loss=0.1827, simple_loss=0.2687, pruned_loss=0.0484, over 1622729.00 frames. +2023-03-20 20:12:38,316 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-20 20:12:44,348 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 20:12:46,486 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19785.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 20:12:51,960 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2139, 2.2902, 1.9099, 3.4488, 1.4882, 3.0488, 1.2099, 3.1506], + device='cuda:1'), covar=tensor([0.0051, 0.0795, 0.2060, 0.0037, 0.4281, 0.0070, 0.1028, 0.0074], + device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0250, 0.0319, 0.0124, 0.0300, 0.0137, 0.0263, 0.0158], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 20:12:54,660 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 20:12:58,031 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.946e+02 2.948e+02 3.559e+02 4.646e+02 7.849e+02, threshold=7.118e+02, percent-clipped=2.0 +2023-03-20 20:13:01,532 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 20:13:03,535 INFO [train.py:901] (1/2) Epoch 8, batch 50, loss[loss=0.21, simple_loss=0.2795, pruned_loss=0.07027, over 7273.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2743, pruned_loss=0.07465, over 327079.92 frames. ], batch size: 68, lr: 1.99e-02, grad_scale: 16.0 +2023-03-20 20:13:03,549 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 20:13:05,378 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1040, 1.9181, 1.9126, 3.2552, 1.5069, 2.8316, 1.2599, 3.1635], + device='cuda:1'), covar=tensor([0.0035, 0.1012, 0.1616, 0.0032, 0.4019, 0.0054, 0.1061, 0.0064], + device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0250, 0.0320, 0.0123, 0.0301, 0.0137, 0.0263, 0.0158], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 20:13:06,445 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5592, 2.7818, 2.2598, 3.6682, 1.5437, 3.1634, 1.6208, 3.4845], + device='cuda:1'), covar=tensor([0.0049, 0.0695, 0.1508, 0.0032, 0.4619, 0.0056, 0.0906, 0.0053], + device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0250, 0.0321, 0.0123, 0.0301, 0.0138, 0.0263, 0.0158], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 20:13:07,314 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 20:13:09,066 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 +2023-03-20 20:13:11,218 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=19833.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 20:13:15,813 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-03-20 20:13:16,318 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-03-20 20:13:24,080 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 20:13:24,559 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 20:13:29,640 INFO [train.py:901] (1/2) Epoch 8, batch 100, loss[loss=0.1909, simple_loss=0.2526, pruned_loss=0.06466, over 7277.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2733, pruned_loss=0.07558, over 574283.53 frames. ], batch size: 66, lr: 1.99e-02, grad_scale: 16.0 +2023-03-20 20:13:49,793 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.917e+02 2.926e+02 3.845e+02 4.661e+02 1.011e+03, threshold=7.690e+02, percent-clipped=6.0 +2023-03-20 20:13:55,386 INFO [train.py:901] (1/2) Epoch 8, batch 150, loss[loss=0.2397, simple_loss=0.2971, pruned_loss=0.09115, over 7263.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2733, pruned_loss=0.07467, over 767441.51 frames. ], batch size: 57, lr: 1.99e-02, grad_scale: 16.0 +2023-03-20 20:14:02,563 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2600, 2.5114, 2.1498, 3.5099, 3.1382, 3.1501, 3.2108, 3.1650], + device='cuda:1'), covar=tensor([0.0969, 0.0406, 0.1449, 0.0176, 0.0028, 0.0026, 0.0035, 0.0030], + device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0211, 0.0261, 0.0197, 0.0110, 0.0106, 0.0114, 0.0110], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 20:14:09,856 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 +2023-03-20 20:14:21,428 INFO [train.py:901] (1/2) Epoch 8, batch 200, loss[loss=0.2579, simple_loss=0.3075, pruned_loss=0.1041, over 6859.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2701, pruned_loss=0.07285, over 915865.07 frames. ], batch size: 107, lr: 1.99e-02, grad_scale: 8.0 +2023-03-20 20:14:24,080 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19974.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:14:25,573 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 20:14:30,035 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 20:14:31,228 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5833, 3.5248, 2.7727, 3.3130, 2.4796, 2.2084, 3.5673, 2.7251], + device='cuda:1'), covar=tensor([0.0088, 0.0110, 0.0247, 0.0063, 0.0301, 0.0406, 0.0144, 0.0460], + device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0212, 0.0224, 0.0199, 0.0285, 0.0289, 0.0233, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 20:14:37,099 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 20:14:46,016 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.800e+02 3.019e+02 3.456e+02 4.343e+02 8.927e+02, threshold=6.913e+02, percent-clipped=3.0 +2023-03-20 20:14:46,699 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4191, 4.2731, 4.2260, 4.7607, 4.7577, 4.7656, 4.1293, 4.3861], + device='cuda:1'), covar=tensor([0.0879, 0.2274, 0.1921, 0.0944, 0.0563, 0.1248, 0.0624, 0.0867], + device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0233, 0.0205, 0.0188, 0.0149, 0.0247, 0.0132, 0.0163], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002], + device='cuda:1') +2023-03-20 20:14:51,100 INFO [train.py:901] (1/2) Epoch 8, batch 250, loss[loss=0.1866, simple_loss=0.247, pruned_loss=0.0631, over 7364.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2696, pruned_loss=0.07217, over 1033155.21 frames. ], batch size: 44, lr: 1.99e-02, grad_scale: 8.0 +2023-03-20 20:14:51,790 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20020.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:14:53,647 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 20:15:07,246 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20050.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:15:11,001 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.01 vs. limit=2.0 +2023-03-20 20:15:13,788 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 20:15:16,847 INFO [train.py:901] (1/2) Epoch 8, batch 300, loss[loss=0.2393, simple_loss=0.2972, pruned_loss=0.09067, over 7226.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2698, pruned_loss=0.07257, over 1123629.93 frames. ], batch size: 93, lr: 1.98e-02, grad_scale: 8.0 +2023-03-20 20:15:22,673 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 20:15:23,794 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20081.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:15:27,745 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20089.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:15:32,214 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=20098.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:15:32,563 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 +2023-03-20 20:15:37,676 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.998e+02 3.160e+02 3.716e+02 4.730e+02 1.016e+03, threshold=7.433e+02, percent-clipped=7.0 +2023-03-20 20:15:39,329 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20112.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:15:43,566 INFO [train.py:901] (1/2) Epoch 8, batch 350, loss[loss=0.2099, simple_loss=0.2717, pruned_loss=0.07407, over 7344.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2692, pruned_loss=0.07246, over 1192049.42 frames. ], batch size: 51, lr: 1.98e-02, grad_scale: 8.0 +2023-03-20 20:15:54,971 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 20:15:57,246 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 20:15:59,141 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20150.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:16:08,565 INFO [train.py:901] (1/2) Epoch 8, batch 400, loss[loss=0.1746, simple_loss=0.2413, pruned_loss=0.05401, over 7160.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2691, pruned_loss=0.07234, over 1246795.07 frames. ], batch size: 41, lr: 1.98e-02, grad_scale: 8.0 +2023-03-20 20:16:11,247 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20173.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:16:22,898 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 +2023-03-20 20:16:27,605 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20204.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:16:29,998 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.659e+02 3.295e+02 4.120e+02 4.990e+02 9.338e+02, threshold=8.240e+02, percent-clipped=6.0 +2023-03-20 20:16:32,696 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4851, 1.1876, 0.9770, 1.1161, 1.3049, 1.2354, 1.4277, 1.0770], + device='cuda:1'), covar=tensor([0.0112, 0.0124, 0.0568, 0.0058, 0.0116, 0.0104, 0.0131, 0.0147], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0017, 0.0016, 0.0015, 0.0018, 0.0018, 0.0018, 0.0019], + device='cuda:1'), out_proj_covar=tensor([2.3691e-05, 2.0426e-05, 2.1834e-05, 1.7190e-05, 2.0962e-05, 2.1484e-05, + 2.2523e-05, 2.6950e-05], device='cuda:1') +2023-03-20 20:16:35,074 INFO [train.py:901] (1/2) Epoch 8, batch 450, loss[loss=0.2161, simple_loss=0.2787, pruned_loss=0.07677, over 7275.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2701, pruned_loss=0.07235, over 1291216.14 frames. ], batch size: 77, lr: 1.98e-02, grad_scale: 8.0 +2023-03-20 20:16:37,541 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 20:16:38,022 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 20:16:58,754 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20265.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:17:00,640 INFO [train.py:901] (1/2) Epoch 8, batch 500, loss[loss=0.2087, simple_loss=0.2647, pruned_loss=0.07636, over 7347.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2693, pruned_loss=0.07205, over 1323716.91 frames. ], batch size: 44, lr: 1.97e-02, grad_scale: 8.0 +2023-03-20 20:17:01,263 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0984, 5.5585, 5.6381, 5.4586, 5.2666, 5.1195, 5.5942, 5.3887], + device='cuda:1'), covar=tensor([0.0286, 0.0294, 0.0396, 0.0398, 0.0311, 0.0277, 0.0342, 0.0475], + device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0139, 0.0111, 0.0108, 0.0092, 0.0133, 0.0116, 0.0096], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 20:17:01,789 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2458, 4.2801, 3.7327, 3.5578, 4.1230, 2.2362, 1.3235, 4.2581], + device='cuda:1'), covar=tensor([0.0021, 0.0020, 0.0076, 0.0071, 0.0027, 0.0466, 0.0687, 0.0037], + device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0051, 0.0075, 0.0060, 0.0066, 0.0095, 0.0104, 0.0064], + device='cuda:1'), out_proj_covar=tensor([7.5829e-05, 7.9897e-05, 1.0798e-04, 8.8854e-05, 8.8881e-05, 1.3904e-04, + 1.5061e-04, 8.9626e-05], device='cuda:1') +2023-03-20 20:17:03,302 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20274.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:17:10,345 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 20:17:12,869 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 20:17:13,386 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 20:17:15,335 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 20:17:19,854 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 20:17:21,307 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.055e+02 2.899e+02 3.508e+02 4.098e+02 7.899e+02, threshold=7.016e+02, percent-clipped=0.0 +2023-03-20 20:17:26,331 INFO [train.py:901] (1/2) Epoch 8, batch 550, loss[loss=0.2009, simple_loss=0.2728, pruned_loss=0.06445, over 7224.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2691, pruned_loss=0.07174, over 1350141.87 frames. ], batch size: 93, lr: 1.97e-02, grad_scale: 8.0 +2023-03-20 20:17:27,863 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=20322.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:17:30,814 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 20:17:38,942 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 20:17:42,622 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 20:17:49,182 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 20:17:51,314 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7450, 2.1418, 2.8327, 2.7957, 2.7722, 2.2130, 2.1561, 2.6451], + device='cuda:1'), covar=tensor([0.1813, 0.0229, 0.2022, 0.1755, 0.0947, 0.2831, 0.2228, 0.1530], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0026, 0.0031, 0.0030, 0.0027, 0.0027, 0.0037, 0.0028], + device='cuda:1'), out_proj_covar=tensor([1.0111e-04, 9.1941e-05, 1.0628e-04, 1.0314e-04, 9.7924e-05, 9.9156e-05, + 1.1910e-04, 1.0163e-04], device='cuda:1') +2023-03-20 20:17:52,182 INFO [train.py:901] (1/2) Epoch 8, batch 600, loss[loss=0.2066, simple_loss=0.2719, pruned_loss=0.07063, over 7334.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2683, pruned_loss=0.07118, over 1371831.16 frames. ], batch size: 49, lr: 1.97e-02, grad_scale: 8.0 +2023-03-20 20:17:56,298 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20376.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:18:02,379 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1941, 4.3194, 3.6657, 3.5370, 4.0720, 2.4780, 1.6793, 4.2766], + device='cuda:1'), covar=tensor([0.0024, 0.0031, 0.0087, 0.0078, 0.0034, 0.0420, 0.0636, 0.0034], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0051, 0.0075, 0.0059, 0.0067, 0.0095, 0.0103, 0.0063], + device='cuda:1'), out_proj_covar=tensor([7.4308e-05, 7.9631e-05, 1.0810e-04, 8.7279e-05, 8.9921e-05, 1.3868e-04, + 1.4974e-04, 8.8868e-05], device='cuda:1') +2023-03-20 20:18:02,404 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20388.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:18:04,789 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 20:18:13,130 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.802e+02 3.123e+02 3.615e+02 4.494e+02 7.871e+02, threshold=7.229e+02, percent-clipped=1.0 +2023-03-20 20:18:13,184 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 20:18:17,979 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20418.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:18:18,361 INFO [train.py:901] (1/2) Epoch 8, batch 650, loss[loss=0.1987, simple_loss=0.2687, pruned_loss=0.06433, over 7336.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2676, pruned_loss=0.0704, over 1386616.31 frames. ], batch size: 61, lr: 1.97e-02, grad_scale: 8.0 +2023-03-20 20:18:27,607 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3374, 0.9719, 0.8242, 1.4810, 1.5292, 1.1178, 1.1806, 1.0871], + device='cuda:1'), covar=tensor([0.0424, 0.1684, 0.0310, 0.0273, 0.0775, 0.1105, 0.0476, 0.0522], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0033, 0.0029, 0.0028, 0.0028, 0.0027, 0.0035, 0.0031], + device='cuda:1'), out_proj_covar=tensor([5.3340e-05, 7.1144e-05, 5.1574e-05, 5.2043e-05, 5.6372e-05, 5.4203e-05, + 6.5097e-05, 6.2324e-05], device='cuda:1') +2023-03-20 20:18:30,956 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 20:18:32,028 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20445.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:18:34,126 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20449.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:18:41,228 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 20:18:44,391 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20468.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:18:44,829 INFO [train.py:901] (1/2) Epoch 8, batch 700, loss[loss=0.2122, simple_loss=0.2758, pruned_loss=0.07428, over 7334.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2675, pruned_loss=0.07042, over 1400034.16 frames. ], batch size: 54, lr: 1.96e-02, grad_scale: 8.0 +2023-03-20 20:18:49,956 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20479.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:18:53,933 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2466, 0.9547, 0.7362, 1.3869, 1.3640, 0.9383, 1.2071, 0.9440], + device='cuda:1'), covar=tensor([0.0602, 0.1416, 0.0569, 0.0585, 0.0895, 0.1510, 0.0623, 0.0587], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0034, 0.0031, 0.0030, 0.0029, 0.0029, 0.0036, 0.0033], + device='cuda:1'), out_proj_covar=tensor([5.4675e-05, 7.3889e-05, 5.4171e-05, 5.4807e-05, 5.9505e-05, 5.7996e-05, + 6.8099e-05, 6.4495e-05], device='cuda:1') +2023-03-20 20:19:03,868 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 20:19:04,389 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 20:19:04,881 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.681e+02 2.952e+02 3.612e+02 4.322e+02 1.243e+03, threshold=7.225e+02, percent-clipped=3.0 +2023-03-20 20:19:10,609 INFO [train.py:901] (1/2) Epoch 8, batch 750, loss[loss=0.2008, simple_loss=0.2622, pruned_loss=0.06968, over 7263.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2682, pruned_loss=0.07066, over 1409271.02 frames. ], batch size: 64, lr: 1.96e-02, grad_scale: 8.0 +2023-03-20 20:19:18,564 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 20:19:23,719 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 20:19:29,799 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 20:19:30,848 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 20:19:31,945 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20560.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:19:33,479 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5037, 3.5442, 3.3406, 3.4773, 3.4798, 3.5574, 3.8789, 3.9627], + device='cuda:1'), covar=tensor([0.0225, 0.0187, 0.0275, 0.0227, 0.0382, 0.0322, 0.0260, 0.0151], + device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0093, 0.0087, 0.0098, 0.0095, 0.0075, 0.0078, 0.0072], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 20:19:35,619 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-03-20 20:19:36,351 INFO [train.py:901] (1/2) Epoch 8, batch 800, loss[loss=0.2057, simple_loss=0.2735, pruned_loss=0.06895, over 7217.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2686, pruned_loss=0.07072, over 1416698.74 frames. ], batch size: 93, lr: 1.96e-02, grad_scale: 8.0 +2023-03-20 20:19:41,464 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 20:19:41,609 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5202, 1.0995, 1.3016, 1.3724, 1.3761, 1.3531, 1.3352, 1.0456], + device='cuda:1'), covar=tensor([0.0161, 0.0091, 0.0294, 0.0064, 0.0102, 0.0113, 0.0106, 0.0246], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0018, 0.0017, 0.0017, 0.0019, 0.0019, 0.0018, 0.0021], + device='cuda:1'), out_proj_covar=tensor([2.6752e-05, 2.1060e-05, 2.2413e-05, 1.8741e-05, 2.1837e-05, 2.2918e-05, + 2.2590e-05, 2.9782e-05], device='cuda:1') +2023-03-20 20:19:44,596 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0322, 3.9916, 3.6649, 3.9869, 3.8039, 4.1132, 3.8810, 3.8641], + device='cuda:1'), covar=tensor([0.0036, 0.0072, 0.0051, 0.0032, 0.0040, 0.0031, 0.0051, 0.0055], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0034, 0.0033, 0.0031, 0.0032, 0.0032, 0.0039, 0.0038], + device='cuda:1'), out_proj_covar=tensor([8.3379e-05, 1.0960e-04, 1.0901e-04, 9.1177e-05, 9.6240e-05, 9.7103e-05, + 1.2508e-04, 1.1772e-04], device='cuda:1') +2023-03-20 20:19:46,141 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20588.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:19:57,366 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.523e+02 2.783e+02 3.356e+02 4.383e+02 1.008e+03, threshold=6.712e+02, percent-clipped=2.0 +2023-03-20 20:20:00,965 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 20:20:01,441 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 20:20:01,873 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.16 vs. limit=5.0 +2023-03-20 20:20:02,462 INFO [train.py:901] (1/2) Epoch 8, batch 850, loss[loss=0.2479, simple_loss=0.3059, pruned_loss=0.09496, over 7224.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2683, pruned_loss=0.07031, over 1422762.84 frames. ], batch size: 93, lr: 1.96e-02, grad_scale: 8.0 +2023-03-20 20:20:07,062 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 20:20:11,162 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 20:20:11,842 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1083, 0.8783, 0.7536, 1.2577, 1.4127, 0.9784, 1.0226, 0.9848], + device='cuda:1'), covar=tensor([0.0876, 0.1657, 0.0803, 0.0847, 0.0362, 0.0458, 0.0395, 0.1087], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0035, 0.0031, 0.0030, 0.0028, 0.0027, 0.0035, 0.0034], + device='cuda:1'), out_proj_covar=tensor([5.3928e-05, 7.3719e-05, 5.4117e-05, 5.4381e-05, 5.7223e-05, 5.5649e-05, + 6.6582e-05, 6.6110e-05], device='cuda:1') +2023-03-20 20:20:17,893 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-03-20 20:20:18,253 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20649.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:20:27,907 INFO [train.py:901] (1/2) Epoch 8, batch 900, loss[loss=0.207, simple_loss=0.2688, pruned_loss=0.0726, over 7357.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2688, pruned_loss=0.07066, over 1428686.24 frames. ], batch size: 61, lr: 1.96e-02, grad_scale: 8.0 +2023-03-20 20:20:31,496 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20676.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:20:47,514 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2407, 1.4688, 1.4441, 1.3058, 1.3273, 1.2847, 1.2648, 1.1751], + device='cuda:1'), covar=tensor([0.0324, 0.0232, 0.0151, 0.0064, 0.0244, 0.0198, 0.0232, 0.0281], + device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0019, 0.0019, 0.0019, 0.0019, 0.0017, 0.0020, 0.0020], + device='cuda:1'), out_proj_covar=tensor([4.6470e-05, 4.3827e-05, 4.2187e-05, 3.6493e-05, 4.4644e-05, 4.0485e-05, + 4.3863e-05, 4.6633e-05], device='cuda:1') +2023-03-20 20:20:48,348 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.807e+02 3.423e+02 4.064e+02 5.075e+02 9.468e+02, threshold=8.127e+02, percent-clipped=9.0 +2023-03-20 20:20:48,368 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 20:20:54,020 INFO [train.py:901] (1/2) Epoch 8, batch 950, loss[loss=0.3025, simple_loss=0.3387, pruned_loss=0.1331, over 6726.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2683, pruned_loss=0.07081, over 1429115.97 frames. ], batch size: 107, lr: 1.95e-02, grad_scale: 8.0 +2023-03-20 20:20:56,544 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=20724.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:21:06,366 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 +2023-03-20 20:21:06,585 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20744.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:21:07,116 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20745.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:21:08,170 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7963, 2.1161, 1.7183, 3.0782, 2.7508, 2.8963, 2.6648, 2.9532], + device='cuda:1'), covar=tensor([0.1356, 0.0729, 0.2111, 0.0250, 0.0050, 0.0069, 0.0048, 0.0056], + device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0208, 0.0260, 0.0205, 0.0108, 0.0110, 0.0109, 0.0112], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 20:21:11,505 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 20:21:18,226 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1356, 2.3576, 2.1556, 3.1943, 2.8230, 3.3809, 2.9314, 3.0753], + device='cuda:1'), covar=tensor([0.0921, 0.0455, 0.1234, 0.0240, 0.0023, 0.0045, 0.0032, 0.0037], + device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0213, 0.0267, 0.0211, 0.0110, 0.0114, 0.0111, 0.0115], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 20:21:18,634 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20768.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:21:19,035 INFO [train.py:901] (1/2) Epoch 8, batch 1000, loss[loss=0.1705, simple_loss=0.2351, pruned_loss=0.05289, over 7147.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2675, pruned_loss=0.07018, over 1430514.05 frames. ], batch size: 41, lr: 1.95e-02, grad_scale: 8.0 +2023-03-20 20:21:19,741 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5044, 3.5654, 2.5052, 3.9008, 2.9937, 3.6710, 2.1601, 2.1696], + device='cuda:1'), covar=tensor([0.0048, 0.0266, 0.0695, 0.0087, 0.0150, 0.0121, 0.1048, 0.0798], + device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0176, 0.0281, 0.0163, 0.0189, 0.0167, 0.0257, 0.0273], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 20:21:21,629 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20774.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:21:31,577 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-03-20 20:21:31,711 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=20793.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:21:32,157 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 20:21:40,617 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 3.000e+02 3.544e+02 4.329e+02 8.832e+02, threshold=7.088e+02, percent-clipped=1.0 +2023-03-20 20:21:41,485 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 20:21:44,204 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=20816.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:21:45,673 INFO [train.py:901] (1/2) Epoch 8, batch 1050, loss[loss=0.2237, simple_loss=0.2823, pruned_loss=0.08255, over 7285.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.267, pruned_loss=0.06973, over 1429927.82 frames. ], batch size: 57, lr: 1.95e-02, grad_scale: 8.0 +2023-03-20 20:21:50,074 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.11 vs. limit=5.0 +2023-03-20 20:21:52,115 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-20 20:21:54,665 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 20:21:58,687 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 20:22:04,442 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6767, 4.8348, 4.4905, 3.9107, 4.6335, 3.1583, 2.2540, 4.6249], + device='cuda:1'), covar=tensor([0.0012, 0.0031, 0.0035, 0.0050, 0.0012, 0.0281, 0.0506, 0.0025], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0052, 0.0076, 0.0059, 0.0069, 0.0095, 0.0103, 0.0064], + device='cuda:1'), out_proj_covar=tensor([7.2053e-05, 8.2100e-05, 1.0957e-04, 8.7708e-05, 9.3243e-05, 1.3709e-04, + 1.4900e-04, 9.1281e-05], device='cuda:1') +2023-03-20 20:22:06,459 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20860.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:22:11,560 INFO [train.py:901] (1/2) Epoch 8, batch 1100, loss[loss=0.2255, simple_loss=0.2837, pruned_loss=0.08358, over 7284.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2666, pruned_loss=0.0696, over 1433978.79 frames. ], batch size: 86, lr: 1.95e-02, grad_scale: 8.0 +2023-03-20 20:22:13,371 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 +2023-03-20 20:22:26,562 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-03-20 20:22:28,658 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 20:22:29,129 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 20:22:31,688 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=20908.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:22:32,090 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.881e+02 3.698e+02 4.714e+02 1.346e+03, threshold=7.395e+02, percent-clipped=2.0 +2023-03-20 20:22:33,012 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-20 20:22:37,180 INFO [train.py:901] (1/2) Epoch 8, batch 1150, loss[loss=0.1906, simple_loss=0.2594, pruned_loss=0.06092, over 7273.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2678, pruned_loss=0.07061, over 1434844.41 frames. ], batch size: 52, lr: 1.94e-02, grad_scale: 8.0 +2023-03-20 20:22:42,252 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 20:22:42,272 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 20:22:42,376 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2727, 2.6900, 3.5086, 2.8177, 3.2528, 2.7121, 2.4750, 3.1170], + device='cuda:1'), covar=tensor([0.1670, 0.0508, 0.1472, 0.2835, 0.1061, 0.1898, 0.2910, 0.1673], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0027, 0.0031, 0.0032, 0.0028, 0.0027, 0.0039, 0.0028], + device='cuda:1'), out_proj_covar=tensor([1.0505e-04, 9.8504e-05, 1.0919e-04, 1.1271e-04, 1.0268e-04, 1.0246e-04, + 1.2686e-04, 1.0624e-04], device='cuda:1') +2023-03-20 20:22:49,927 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20944.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:23:02,934 INFO [train.py:901] (1/2) Epoch 8, batch 1200, loss[loss=0.1502, simple_loss=0.2114, pruned_loss=0.0445, over 6974.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2672, pruned_loss=0.07012, over 1436505.65 frames. ], batch size: 35, lr: 1.94e-02, grad_scale: 8.0 +2023-03-20 20:23:16,161 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 20:23:24,114 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.032e+02 2.800e+02 3.354e+02 4.128e+02 7.316e+02, threshold=6.709e+02, percent-clipped=0.0 +2023-03-20 20:23:25,496 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-03-20 20:23:28,991 INFO [train.py:901] (1/2) Epoch 8, batch 1250, loss[loss=0.1657, simple_loss=0.2309, pruned_loss=0.05024, over 7188.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2679, pruned_loss=0.0705, over 1438441.60 frames. ], batch size: 39, lr: 1.94e-02, grad_scale: 8.0 +2023-03-20 20:23:39,944 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 20:23:42,174 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21044.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:23:44,470 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 20:23:45,472 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 20:23:46,133 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21052.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:23:54,570 INFO [train.py:901] (1/2) Epoch 8, batch 1300, loss[loss=0.1996, simple_loss=0.2639, pruned_loss=0.06766, over 7318.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2692, pruned_loss=0.07156, over 1439943.32 frames. ], batch size: 59, lr: 1.94e-02, grad_scale: 8.0 +2023-03-20 20:23:57,720 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21074.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:24:06,773 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=21092.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:24:08,782 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 20:24:11,222 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 20:24:12,042 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-03-20 20:24:13,816 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 20:24:15,289 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 3.023e+02 3.781e+02 4.755e+02 9.380e+02, threshold=7.562e+02, percent-clipped=5.0 +2023-03-20 20:24:17,463 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21113.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:24:19,982 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21118.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:24:20,340 INFO [train.py:901] (1/2) Epoch 8, batch 1350, loss[loss=0.2, simple_loss=0.261, pruned_loss=0.06954, over 7330.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2701, pruned_loss=0.07193, over 1441676.54 frames. ], batch size: 44, lr: 1.94e-02, grad_scale: 8.0 +2023-03-20 20:24:21,900 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=21122.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:24:23,883 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 20:24:46,753 INFO [train.py:901] (1/2) Epoch 8, batch 1400, loss[loss=0.2134, simple_loss=0.2755, pruned_loss=0.07571, over 7362.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2685, pruned_loss=0.07147, over 1439063.84 frames. ], batch size: 73, lr: 1.93e-02, grad_scale: 8.0 +2023-03-20 20:24:51,905 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21179.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:24:58,773 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 20:25:07,013 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.034e+02 2.789e+02 3.546e+02 4.120e+02 7.896e+02, threshold=7.092e+02, percent-clipped=1.0 +2023-03-20 20:25:09,638 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8154, 2.6717, 3.1397, 2.8562, 3.1382, 2.0714, 2.2878, 2.7746], + device='cuda:1'), covar=tensor([0.1586, 0.0348, 0.1068, 0.2265, 0.0711, 0.2434, 0.2352, 0.1800], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0027, 0.0030, 0.0030, 0.0029, 0.0026, 0.0038, 0.0028], + device='cuda:1'), out_proj_covar=tensor([1.0371e-04, 9.7774e-05, 1.0680e-04, 1.0915e-04, 1.0445e-04, 1.0103e-04, + 1.2457e-04, 1.0585e-04], device='cuda:1') +2023-03-20 20:25:11,978 INFO [train.py:901] (1/2) Epoch 8, batch 1450, loss[loss=0.2244, simple_loss=0.2841, pruned_loss=0.08241, over 7306.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2685, pruned_loss=0.07147, over 1441649.82 frames. ], batch size: 75, lr: 1.93e-02, grad_scale: 8.0 +2023-03-20 20:25:12,695 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8610, 2.1354, 2.0705, 3.1832, 3.0666, 3.1999, 3.0205, 2.8095], + device='cuda:1'), covar=tensor([0.1080, 0.0500, 0.1410, 0.0245, 0.0058, 0.0050, 0.0031, 0.0031], + device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0216, 0.0271, 0.0217, 0.0108, 0.0114, 0.0114, 0.0116], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 20:25:15,365 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-03-20 20:25:19,382 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9822, 3.9251, 3.3598, 3.2826, 3.7397, 2.3992, 1.6232, 3.8152], + device='cuda:1'), covar=tensor([0.0014, 0.0050, 0.0094, 0.0065, 0.0039, 0.0367, 0.0553, 0.0042], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0053, 0.0078, 0.0061, 0.0071, 0.0095, 0.0104, 0.0066], + device='cuda:1'), out_proj_covar=tensor([6.9666e-05, 8.2539e-05, 1.1330e-04, 9.0134e-05, 9.5377e-05, 1.3698e-04, + 1.4971e-04, 9.3213e-05], device='cuda:1') +2023-03-20 20:25:22,295 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 20:25:25,332 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21244.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:25:36,021 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2149, 2.5989, 2.1789, 2.2215, 2.5022, 2.0669, 2.4198, 2.0319], + device='cuda:1'), covar=tensor([0.0896, 0.0323, 0.0876, 0.0999, 0.1024, 0.0681, 0.1536, 0.1317], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0032, 0.0040, 0.0035, 0.0037, 0.0037, 0.0034, 0.0035], + device='cuda:1'), out_proj_covar=tensor([1.2013e-04, 9.7375e-05, 1.1235e-04, 1.0152e-04, 1.0778e-04, 1.0753e-04, + 1.0068e-04, 1.0181e-04], device='cuda:1') +2023-03-20 20:25:37,951 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 20:25:38,427 INFO [train.py:901] (1/2) Epoch 8, batch 1500, loss[loss=0.1848, simple_loss=0.2571, pruned_loss=0.05618, over 7279.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2684, pruned_loss=0.07099, over 1442595.78 frames. ], batch size: 66, lr: 1.93e-02, grad_scale: 8.0 +2023-03-20 20:25:41,639 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0777, 2.2877, 2.2235, 3.2186, 1.4968, 3.0452, 1.1755, 3.3340], + device='cuda:1'), covar=tensor([0.0051, 0.0968, 0.1594, 0.0039, 0.4472, 0.0066, 0.1187, 0.0138], + device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0268, 0.0330, 0.0130, 0.0318, 0.0143, 0.0283, 0.0173], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0001, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 20:25:46,180 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1894, 3.5617, 3.8818, 3.8003, 3.7879, 3.7149, 4.0649, 3.7665], + device='cuda:1'), covar=tensor([0.0080, 0.0151, 0.0120, 0.0109, 0.0137, 0.0091, 0.0115, 0.0110], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0049, 0.0055, 0.0042, 0.0068, 0.0058, 0.0053, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 20:25:50,163 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=21292.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:25:59,227 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 2.932e+02 3.502e+02 4.369e+02 7.358e+02, threshold=7.004e+02, percent-clipped=1.0 +2023-03-20 20:26:00,899 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21312.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:26:01,774 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 20:26:04,318 INFO [train.py:901] (1/2) Epoch 8, batch 1550, loss[loss=0.1942, simple_loss=0.2647, pruned_loss=0.06187, over 7339.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2679, pruned_loss=0.07072, over 1442771.39 frames. ], batch size: 63, lr: 1.93e-02, grad_scale: 8.0 +2023-03-20 20:26:08,915 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21328.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:26:25,301 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 +2023-03-20 20:26:30,049 INFO [train.py:901] (1/2) Epoch 8, batch 1600, loss[loss=0.1945, simple_loss=0.2658, pruned_loss=0.0616, over 7237.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2681, pruned_loss=0.07092, over 1441160.71 frames. ], batch size: 55, lr: 1.93e-02, grad_scale: 8.0 +2023-03-20 20:26:32,179 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21373.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:26:33,050 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 20:26:34,116 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 20:26:36,685 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 20:26:40,336 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21389.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:26:42,317 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5754, 3.4588, 3.1830, 3.4785, 3.2879, 3.4374, 3.2808, 3.2923], + device='cuda:1'), covar=tensor([0.0036, 0.0067, 0.0050, 0.0038, 0.0048, 0.0035, 0.0062, 0.0065], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0033, 0.0031, 0.0029, 0.0031, 0.0031, 0.0038, 0.0036], + device='cuda:1'), out_proj_covar=tensor([8.2968e-05, 1.0653e-04, 1.0167e-04, 8.3876e-05, 9.6230e-05, 9.2658e-05, + 1.2357e-04, 1.1046e-04], device='cuda:1') +2023-03-20 20:26:47,096 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 20:26:50,767 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21408.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:26:51,134 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.861e+02 3.255e+02 3.633e+02 4.410e+02 1.218e+03, threshold=7.266e+02, percent-clipped=6.0 +2023-03-20 20:26:51,157 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 20:26:56,167 INFO [train.py:901] (1/2) Epoch 8, batch 1650, loss[loss=0.2043, simple_loss=0.2731, pruned_loss=0.06779, over 7388.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2669, pruned_loss=0.07011, over 1440605.54 frames. ], batch size: 65, lr: 1.92e-02, grad_scale: 8.0 +2023-03-20 20:27:00,390 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 20:27:15,913 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 20:27:16,692 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 +2023-03-20 20:27:20,397 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 20:27:21,865 INFO [train.py:901] (1/2) Epoch 8, batch 1700, loss[loss=0.2037, simple_loss=0.2704, pruned_loss=0.06853, over 7327.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.267, pruned_loss=0.06966, over 1443417.65 frames. ], batch size: 49, lr: 1.92e-02, grad_scale: 8.0 +2023-03-20 20:27:22,076 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1957, 3.9819, 3.5404, 3.8364, 3.0733, 2.8330, 4.1221, 3.2504], + device='cuda:1'), covar=tensor([0.0074, 0.0101, 0.0129, 0.0072, 0.0238, 0.0311, 0.0143, 0.0435], + device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0222, 0.0224, 0.0210, 0.0283, 0.0283, 0.0235, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 20:27:24,459 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21474.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:27:31,035 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 20:27:35,215 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0770, 2.4021, 2.1717, 3.3952, 3.4737, 3.2613, 3.0311, 3.0872], + device='cuda:1'), covar=tensor([0.1112, 0.0566, 0.1627, 0.0312, 0.0100, 0.0039, 0.0089, 0.0040], + device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0210, 0.0265, 0.0215, 0.0108, 0.0110, 0.0112, 0.0116], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 20:27:43,084 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.770e+02 2.755e+02 3.361e+02 4.411e+02 8.082e+02, threshold=6.722e+02, percent-clipped=3.0 +2023-03-20 20:27:48,225 INFO [train.py:901] (1/2) Epoch 8, batch 1750, loss[loss=0.1887, simple_loss=0.2615, pruned_loss=0.05796, over 7284.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2674, pruned_loss=0.06954, over 1444611.32 frames. ], batch size: 68, lr: 1.92e-02, grad_scale: 8.0 +2023-03-20 20:27:56,273 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 20:27:57,292 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 20:28:04,953 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4902, 3.3428, 2.7175, 3.2101, 2.5499, 2.3154, 3.3853, 2.6385], + device='cuda:1'), covar=tensor([0.0153, 0.0141, 0.0168, 0.0101, 0.0247, 0.0350, 0.0190, 0.0386], + device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0227, 0.0225, 0.0213, 0.0285, 0.0287, 0.0239, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 20:28:13,427 INFO [train.py:901] (1/2) Epoch 8, batch 1800, loss[loss=0.1965, simple_loss=0.2647, pruned_loss=0.06415, over 7229.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2672, pruned_loss=0.06934, over 1445584.82 frames. ], batch size: 93, lr: 1.92e-02, grad_scale: 8.0 +2023-03-20 20:28:19,536 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 20:28:27,755 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21595.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:28:32,490 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 20:28:34,963 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.818e+02 2.908e+02 3.364e+02 4.556e+02 1.018e+03, threshold=6.727e+02, percent-clipped=4.0 +2023-03-20 20:28:37,150 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21613.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:28:40,032 INFO [train.py:901] (1/2) Epoch 8, batch 1850, loss[loss=0.1998, simple_loss=0.2622, pruned_loss=0.06873, over 7281.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2666, pruned_loss=0.06905, over 1445059.62 frames. ], batch size: 77, lr: 1.91e-02, grad_scale: 8.0 +2023-03-20 20:28:42,549 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 20:28:45,337 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.04 vs. limit=2.0 +2023-03-20 20:28:58,829 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 20:28:59,473 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21656.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:29:05,414 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21668.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:29:05,842 INFO [train.py:901] (1/2) Epoch 8, batch 1900, loss[loss=0.2097, simple_loss=0.2749, pruned_loss=0.07229, over 7310.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2665, pruned_loss=0.06892, over 1443686.84 frames. ], batch size: 83, lr: 1.91e-02, grad_scale: 8.0 +2023-03-20 20:29:09,318 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21674.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:29:14,249 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21684.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:29:23,716 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 20:29:26,323 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21708.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:29:26,688 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.930e+02 2.999e+02 3.585e+02 4.195e+02 1.317e+03, threshold=7.171e+02, percent-clipped=5.0 +2023-03-20 20:29:31,651 INFO [train.py:901] (1/2) Epoch 8, batch 1950, loss[loss=0.2121, simple_loss=0.2692, pruned_loss=0.07752, over 7262.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2666, pruned_loss=0.06916, over 1443166.78 frames. ], batch size: 47, lr: 1.91e-02, grad_scale: 8.0 +2023-03-20 20:29:35,052 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 20:29:39,579 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 20:29:40,113 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 20:29:50,845 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=21756.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:29:51,644 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.77 vs. limit=5.0 +2023-03-20 20:29:55,092 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21763.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:29:58,001 INFO [train.py:901] (1/2) Epoch 8, batch 2000, loss[loss=0.1999, simple_loss=0.2575, pruned_loss=0.07112, over 7238.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2668, pruned_loss=0.06962, over 1442883.81 frames. ], batch size: 45, lr: 1.91e-02, grad_scale: 8.0 +2023-03-20 20:29:58,586 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 20:30:00,716 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21774.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:30:09,241 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 20:30:12,904 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5671, 3.3183, 3.2758, 3.3534, 3.3545, 3.5055, 3.2972, 3.3688], + device='cuda:1'), covar=tensor([0.0029, 0.0076, 0.0047, 0.0044, 0.0044, 0.0039, 0.0070, 0.0059], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0033, 0.0031, 0.0029, 0.0031, 0.0030, 0.0038, 0.0035], + device='cuda:1'), out_proj_covar=tensor([7.9111e-05, 1.0531e-04, 1.0110e-04, 8.3706e-05, 9.2790e-05, 9.0230e-05, + 1.2451e-04, 1.0622e-04], device='cuda:1') +2023-03-20 20:30:14,258 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0719, 3.9659, 3.5080, 3.4999, 3.7717, 2.4475, 1.8823, 4.0022], + device='cuda:1'), covar=tensor([0.0018, 0.0044, 0.0073, 0.0074, 0.0052, 0.0377, 0.0584, 0.0042], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0055, 0.0079, 0.0065, 0.0073, 0.0097, 0.0108, 0.0065], + device='cuda:1'), out_proj_covar=tensor([7.0204e-05, 8.5272e-05, 1.1539e-04, 9.5656e-05, 9.9357e-05, 1.4055e-04, + 1.5550e-04, 9.2074e-05], device='cuda:1') +2023-03-20 20:30:17,661 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 20:30:18,631 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.703e+02 2.773e+02 3.531e+02 4.246e+02 7.839e+02, threshold=7.062e+02, percent-clipped=2.0 +2023-03-20 20:30:23,664 INFO [train.py:901] (1/2) Epoch 8, batch 2050, loss[loss=0.2355, simple_loss=0.2874, pruned_loss=0.09182, over 7287.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.267, pruned_loss=0.06967, over 1443848.98 frames. ], batch size: 68, lr: 1.91e-02, grad_scale: 8.0 +2023-03-20 20:30:25,013 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.02 vs. limit=2.0 +2023-03-20 20:30:25,217 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=21822.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:30:26,310 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21824.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:30:36,377 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6384, 3.2453, 3.2306, 3.2117, 3.3073, 3.3539, 3.4162, 3.3191], + device='cuda:1'), covar=tensor([0.0090, 0.0179, 0.0181, 0.0213, 0.0176, 0.0099, 0.0173, 0.0116], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0049, 0.0054, 0.0044, 0.0068, 0.0056, 0.0054, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 20:30:40,546 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21850.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:30:50,071 INFO [train.py:901] (1/2) Epoch 8, batch 2100, loss[loss=0.2024, simple_loss=0.2752, pruned_loss=0.06478, over 7277.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2658, pruned_loss=0.06894, over 1441399.26 frames. ], batch size: 77, lr: 1.90e-02, grad_scale: 8.0 +2023-03-20 20:30:51,615 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 20:30:54,181 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 20:31:10,070 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.845e+02 2.957e+02 3.693e+02 4.349e+02 9.898e+02, threshold=7.385e+02, percent-clipped=1.0 +2023-03-20 20:31:11,249 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21911.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:31:13,723 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.8506, 5.3445, 5.3922, 5.2391, 4.9434, 4.8475, 5.3775, 5.0866], + device='cuda:1'), covar=tensor([0.0316, 0.0302, 0.0341, 0.0426, 0.0368, 0.0292, 0.0313, 0.0512], + device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0149, 0.0112, 0.0110, 0.0097, 0.0138, 0.0121, 0.0100], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 20:31:15,082 INFO [train.py:901] (1/2) Epoch 8, batch 2150, loss[loss=0.2113, simple_loss=0.2685, pruned_loss=0.07704, over 7269.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2658, pruned_loss=0.06881, over 1441846.13 frames. ], batch size: 47, lr: 1.90e-02, grad_scale: 8.0 +2023-03-20 20:31:32,736 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21951.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:31:40,665 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 20:31:41,252 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21968.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:31:41,636 INFO [train.py:901] (1/2) Epoch 8, batch 2200, loss[loss=0.2104, simple_loss=0.2725, pruned_loss=0.0741, over 7218.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2663, pruned_loss=0.06893, over 1442503.23 frames. ], batch size: 93, lr: 1.90e-02, grad_scale: 16.0 +2023-03-20 20:31:41,712 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21969.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:31:42,265 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21970.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:31:42,797 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6556, 1.0603, 1.8913, 1.8248, 1.6123, 1.2939, 1.2792, 1.1350], + device='cuda:1'), covar=tensor([0.0278, 0.0559, 0.0202, 0.0182, 0.0236, 0.0195, 0.0195, 0.0220], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0018, 0.0017, 0.0018, 0.0020, 0.0018, 0.0018, 0.0021], + device='cuda:1'), out_proj_covar=tensor([2.3110e-05, 2.1471e-05, 2.1872e-05, 1.9996e-05, 2.3114e-05, 2.0903e-05, + 2.1727e-05, 2.8003e-05], device='cuda:1') +2023-03-20 20:31:49,432 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21984.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:32:02,868 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.683e+02 2.848e+02 3.645e+02 4.438e+02 1.724e+03, threshold=7.290e+02, percent-clipped=4.0 +2023-03-20 20:32:06,492 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=22016.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:32:07,935 INFO [train.py:901] (1/2) Epoch 8, batch 2250, loss[loss=0.2076, simple_loss=0.2735, pruned_loss=0.0708, over 7268.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2661, pruned_loss=0.06873, over 1441173.45 frames. ], batch size: 64, lr: 1.90e-02, grad_scale: 16.0 +2023-03-20 20:32:14,731 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22031.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:32:15,129 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=22032.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:32:15,569 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 20:32:16,098 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 20:32:28,684 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 20:32:33,841 INFO [train.py:901] (1/2) Epoch 8, batch 2300, loss[loss=0.1661, simple_loss=0.2356, pruned_loss=0.04831, over 7214.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2655, pruned_loss=0.06854, over 1440079.27 frames. ], batch size: 45, lr: 1.90e-02, grad_scale: 16.0 +2023-03-20 20:32:55,478 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.018e+02 3.001e+02 3.476e+02 4.126e+02 8.245e+02, threshold=6.952e+02, percent-clipped=1.0 +2023-03-20 20:33:00,578 INFO [train.py:901] (1/2) Epoch 8, batch 2350, loss[loss=0.1997, simple_loss=0.2464, pruned_loss=0.07648, over 5881.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2658, pruned_loss=0.06862, over 1439997.35 frames. ], batch size: 26, lr: 1.89e-02, grad_scale: 16.0 +2023-03-20 20:33:00,668 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22119.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:33:06,210 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22130.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:33:14,832 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5554, 3.9568, 4.4039, 4.2761, 4.2120, 4.2051, 4.5242, 4.1883], + device='cuda:1'), covar=tensor([0.0109, 0.0115, 0.0081, 0.0097, 0.0108, 0.0080, 0.0135, 0.0089], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0050, 0.0054, 0.0046, 0.0070, 0.0057, 0.0054, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 20:33:15,779 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 20:33:21,844 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 20:33:25,868 INFO [train.py:901] (1/2) Epoch 8, batch 2400, loss[loss=0.1591, simple_loss=0.2235, pruned_loss=0.04736, over 7174.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2652, pruned_loss=0.06816, over 1438915.80 frames. ], batch size: 39, lr: 1.89e-02, grad_scale: 16.0 +2023-03-20 20:33:29,433 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22176.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:33:32,495 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 20:33:33,340 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-03-20 20:33:35,113 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 20:33:38,327 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22191.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:33:41,977 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-03-20 20:33:46,129 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22206.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:33:47,538 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 2.682e+02 3.425e+02 4.426e+02 1.107e+03, threshold=6.849e+02, percent-clipped=3.0 +2023-03-20 20:33:50,724 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22215.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:33:52,640 INFO [train.py:901] (1/2) Epoch 8, batch 2450, loss[loss=0.2486, simple_loss=0.2966, pruned_loss=0.1003, over 6810.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2659, pruned_loss=0.06876, over 1437491.48 frames. ], batch size: 107, lr: 1.89e-02, grad_scale: 16.0 +2023-03-20 20:34:01,894 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:34:01,908 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:34:02,276 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 20:34:08,828 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22251.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:34:18,446 INFO [train.py:901] (1/2) Epoch 8, batch 2500, loss[loss=0.1979, simple_loss=0.2666, pruned_loss=0.06464, over 7320.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2658, pruned_loss=0.06857, over 1440022.92 frames. ], batch size: 59, lr: 1.89e-02, grad_scale: 16.0 +2023-03-20 20:34:18,554 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22269.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:34:22,653 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22276.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:34:28,379 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 20:34:33,427 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22298.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:34:33,840 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=22299.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:34:38,860 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.527e+02 2.963e+02 3.499e+02 4.353e+02 8.938e+02, threshold=6.998e+02, percent-clipped=3.0 +2023-03-20 20:34:42,937 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=22317.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:34:43,841 INFO [train.py:901] (1/2) Epoch 8, batch 2550, loss[loss=0.1932, simple_loss=0.2531, pruned_loss=0.06664, over 7224.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2658, pruned_loss=0.06857, over 1440269.92 frames. ], batch size: 45, lr: 1.89e-02, grad_scale: 16.0 +2023-03-20 20:34:47,421 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22326.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:35:08,740 INFO [train.py:901] (1/2) Epoch 8, batch 2600, loss[loss=0.1928, simple_loss=0.2488, pruned_loss=0.0684, over 7349.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2664, pruned_loss=0.06869, over 1440795.98 frames. ], batch size: 61, lr: 1.88e-02, grad_scale: 16.0 +2023-03-20 20:35:17,997 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22385.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:35:30,556 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.811e+02 2.880e+02 3.559e+02 4.334e+02 1.074e+03, threshold=7.118e+02, percent-clipped=5.0 +2023-03-20 20:35:35,473 INFO [train.py:901] (1/2) Epoch 8, batch 2650, loss[loss=0.2216, simple_loss=0.2809, pruned_loss=0.08115, over 7283.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2662, pruned_loss=0.06898, over 1442592.75 frames. ], batch size: 68, lr: 1.88e-02, grad_scale: 16.0 +2023-03-20 20:35:35,594 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22419.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:35:36,580 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1488, 4.0594, 3.5941, 3.8059, 2.8964, 2.6597, 4.1441, 3.2397], + device='cuda:1'), covar=tensor([0.0092, 0.0077, 0.0122, 0.0076, 0.0273, 0.0332, 0.0096, 0.0460], + device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0231, 0.0231, 0.0225, 0.0291, 0.0289, 0.0251, 0.0297], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 20:35:48,902 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22446.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:35:52,841 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6540, 3.7082, 2.5847, 3.9303, 3.4540, 3.7511, 2.4071, 2.2571], + device='cuda:1'), covar=tensor([0.0070, 0.0248, 0.0781, 0.0103, 0.0172, 0.0112, 0.0994, 0.0849], + device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0178, 0.0285, 0.0174, 0.0190, 0.0173, 0.0260, 0.0269], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 20:35:59,234 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=22467.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:35:59,516 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 +2023-03-20 20:36:00,192 INFO [train.py:901] (1/2) Epoch 8, batch 2700, loss[loss=0.2047, simple_loss=0.2629, pruned_loss=0.0733, over 7219.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2663, pruned_loss=0.06917, over 1442626.39 frames. ], batch size: 45, lr: 1.88e-02, grad_scale: 16.0 +2023-03-20 20:36:05,898 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7860, 3.9287, 2.6093, 4.1286, 3.6933, 3.9434, 2.5747, 2.2668], + device='cuda:1'), covar=tensor([0.0054, 0.0335, 0.0884, 0.0096, 0.0077, 0.0110, 0.0971, 0.0951], + device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0180, 0.0287, 0.0176, 0.0193, 0.0176, 0.0264, 0.0271], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 20:36:08,761 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22486.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:36:18,521 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22506.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:36:19,868 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.942e+02 2.708e+02 3.424e+02 4.354e+02 9.279e+02, threshold=6.847e+02, percent-clipped=4.0 +2023-03-20 20:36:24,784 INFO [train.py:901] (1/2) Epoch 8, batch 2750, loss[loss=0.1873, simple_loss=0.2541, pruned_loss=0.06025, over 7245.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2661, pruned_loss=0.06864, over 1444269.84 frames. ], batch size: 47, lr: 1.88e-02, grad_scale: 16.0 +2023-03-20 20:36:31,217 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22532.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:36:31,812 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7606, 3.6314, 3.2557, 3.4370, 2.8404, 2.4906, 3.7708, 3.0051], + device='cuda:1'), covar=tensor([0.0139, 0.0089, 0.0131, 0.0092, 0.0240, 0.0361, 0.0163, 0.0428], + device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0230, 0.0234, 0.0228, 0.0292, 0.0296, 0.0252, 0.0303], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 20:36:41,926 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=22554.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:36:47,950 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1920, 2.5250, 1.9736, 2.4524, 2.3199, 2.0575, 2.3480, 2.1478], + device='cuda:1'), covar=tensor([0.0782, 0.0499, 0.0757, 0.0651, 0.0783, 0.0521, 0.0705, 0.1302], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0033, 0.0038, 0.0034, 0.0035, 0.0035, 0.0034, 0.0033], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 20:36:49,289 INFO [train.py:901] (1/2) Epoch 8, batch 2800, loss[loss=0.1515, simple_loss=0.2122, pruned_loss=0.04544, over 6979.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2658, pruned_loss=0.06855, over 1445228.33 frames. ], batch size: 35, lr: 1.88e-02, grad_scale: 16.0 +2023-03-20 20:36:50,262 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22571.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:37:15,245 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 20:37:15,296 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 20:37:15,320 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 20:37:15,336 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 20:37:15,833 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 20:37:15,855 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 20:37:15,866 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 20:37:15,963 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 20:37:16,037 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 20:37:16,053 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 20:37:16,069 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 20:37:16,227 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 20:37:16,232 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 20:37:16,239 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 20:37:16,375 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 20:37:16,378 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 20:37:16,882 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 20:37:23,759 INFO [train.py:901] (1/2) Epoch 9, batch 0, loss[loss=0.2083, simple_loss=0.2747, pruned_loss=0.07102, over 7289.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2747, pruned_loss=0.07102, over 7289.00 frames. ], batch size: 68, lr: 1.80e-02, grad_scale: 16.0 +2023-03-20 20:37:23,759 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 20:37:32,161 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4188, 3.6733, 3.4505, 3.4785, 3.3182, 3.4527, 3.9028, 3.8680], + device='cuda:1'), covar=tensor([0.0263, 0.0188, 0.0211, 0.0245, 0.0408, 0.0250, 0.0206, 0.0196], + device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0096, 0.0090, 0.0101, 0.0096, 0.0081, 0.0078, 0.0077], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 20:37:48,979 INFO [train.py:935] (1/2) Epoch 9, validation: loss=0.1789, simple_loss=0.2654, pruned_loss=0.04614, over 1622729.00 frames. +2023-03-20 20:37:48,980 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-20 20:37:49,083 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22593.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:37:55,451 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 20:37:58,049 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.041e+02 3.046e+02 3.484e+02 4.247e+02 9.512e+02, threshold=6.968e+02, percent-clipped=1.0 +2023-03-20 20:38:07,321 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 20:38:07,388 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22626.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:38:13,650 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22639.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:38:14,524 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 20:38:15,528 INFO [train.py:901] (1/2) Epoch 9, batch 50, loss[loss=0.2052, simple_loss=0.2744, pruned_loss=0.068, over 7366.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2648, pruned_loss=0.06834, over 326141.37 frames. ], batch size: 54, lr: 1.80e-02, grad_scale: 16.0 +2023-03-20 20:38:17,014 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 20:38:19,589 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 20:38:31,160 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=22674.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:38:35,612 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 20:38:35,626 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 20:38:41,215 INFO [train.py:901] (1/2) Epoch 9, batch 100, loss[loss=0.1876, simple_loss=0.2534, pruned_loss=0.0609, over 7295.00 frames. ], tot_loss[loss=0.2, simple_loss=0.265, pruned_loss=0.06752, over 573522.91 frames. ], batch size: 68, lr: 1.80e-02, grad_scale: 16.0 +2023-03-20 20:38:44,924 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22700.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 20:38:45,347 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2233, 4.7258, 4.7937, 4.6129, 4.6538, 4.2824, 4.8183, 4.5812], + device='cuda:1'), covar=tensor([0.0418, 0.0390, 0.0377, 0.0549, 0.0339, 0.0288, 0.0296, 0.0588], + device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0147, 0.0113, 0.0109, 0.0094, 0.0138, 0.0120, 0.0099], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 20:38:45,442 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2971, 3.4554, 2.2149, 3.6246, 2.8902, 3.5054, 2.1074, 1.9106], + device='cuda:1'), covar=tensor([0.0065, 0.0271, 0.0997, 0.0173, 0.0149, 0.0109, 0.1337, 0.1010], + device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0182, 0.0298, 0.0184, 0.0200, 0.0182, 0.0271, 0.0280], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 20:38:49,874 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.903e+02 2.746e+02 3.605e+02 4.493e+02 9.955e+02, threshold=7.209e+02, percent-clipped=4.0 +2023-03-20 20:38:52,705 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 +2023-03-20 20:39:05,886 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22741.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:39:06,856 INFO [train.py:901] (1/2) Epoch 9, batch 150, loss[loss=0.1979, simple_loss=0.2614, pruned_loss=0.06713, over 7270.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2657, pruned_loss=0.0679, over 764975.33 frames. ], batch size: 52, lr: 1.79e-02, grad_scale: 16.0 +2023-03-20 20:39:29,012 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22786.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:39:32,388 INFO [train.py:901] (1/2) Epoch 9, batch 200, loss[loss=0.2277, simple_loss=0.2884, pruned_loss=0.0835, over 7314.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2636, pruned_loss=0.0669, over 918065.30 frames. ], batch size: 83, lr: 1.79e-02, grad_scale: 16.0 +2023-03-20 20:39:33,023 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22794.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:39:38,440 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 20:39:41,346 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.893e+02 3.607e+02 4.516e+02 8.593e+02, threshold=7.215e+02, percent-clipped=2.0 +2023-03-20 20:39:42,894 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 20:39:48,382 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 20:39:49,493 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22825.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:39:52,932 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22832.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:39:53,932 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=22834.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:39:58,394 INFO [train.py:901] (1/2) Epoch 9, batch 250, loss[loss=0.1682, simple_loss=0.2349, pruned_loss=0.05071, over 7329.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2633, pruned_loss=0.0665, over 1034941.01 frames. ], batch size: 44, lr: 1.79e-02, grad_scale: 16.0 +2023-03-20 20:40:01,942 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 20:40:04,545 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22855.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:40:08,190 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1650, 1.2321, 1.2518, 1.2154, 1.2860, 0.8205, 0.9305, 0.9061], + device='cuda:1'), covar=tensor([0.0130, 0.0127, 0.0185, 0.0087, 0.0107, 0.0136, 0.0153, 0.0158], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0018, 0.0016, 0.0018, 0.0020, 0.0018, 0.0019, 0.0021], + device='cuda:1'), out_proj_covar=tensor([2.3639e-05, 2.0518e-05, 2.1303e-05, 2.0197e-05, 2.3063e-05, 2.0983e-05, + 2.1948e-05, 2.8263e-05], device='cuda:1') +2023-03-20 20:40:13,185 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22871.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:40:17,630 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=22880.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:40:21,388 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22886.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:40:24,272 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 20:40:24,761 INFO [train.py:901] (1/2) Epoch 9, batch 300, loss[loss=0.2117, simple_loss=0.2752, pruned_loss=0.07414, over 7120.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2629, pruned_loss=0.06623, over 1125158.98 frames. ], batch size: 98, lr: 1.79e-02, grad_scale: 16.0 +2023-03-20 20:40:24,861 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22893.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:40:32,773 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.746e+02 2.565e+02 3.158e+02 3.848e+02 7.553e+02, threshold=6.316e+02, percent-clipped=2.0 +2023-03-20 20:40:33,319 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 20:40:37,824 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=22919.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:40:48,678 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=22941.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:40:49,527 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.18 vs. limit=5.0 +2023-03-20 20:40:49,619 INFO [train.py:901] (1/2) Epoch 9, batch 350, loss[loss=0.2406, simple_loss=0.3028, pruned_loss=0.08922, over 6701.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2636, pruned_loss=0.06635, over 1196306.81 frames. ], batch size: 107, lr: 1.79e-02, grad_scale: 16.0 +2023-03-20 20:41:03,013 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 +2023-03-20 20:41:07,853 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 20:41:15,826 INFO [train.py:901] (1/2) Epoch 9, batch 400, loss[loss=0.1658, simple_loss=0.2235, pruned_loss=0.05406, over 7053.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2625, pruned_loss=0.06601, over 1251224.11 frames. ], batch size: 35, lr: 1.79e-02, grad_scale: 16.0 +2023-03-20 20:41:16,910 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22995.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 20:41:17,422 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0649, 4.1987, 4.0333, 4.1698, 3.8007, 4.2156, 4.5859, 4.6171], + device='cuda:1'), covar=tensor([0.0188, 0.0157, 0.0188, 0.0174, 0.0487, 0.0188, 0.0211, 0.0171], + device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0095, 0.0090, 0.0101, 0.0098, 0.0079, 0.0078, 0.0076], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 20:41:23,491 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-20 20:41:24,105 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.024e+02 2.885e+02 3.630e+02 4.346e+02 7.423e+02, threshold=7.259e+02, percent-clipped=3.0 +2023-03-20 20:41:28,761 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0899, 2.1062, 1.9401, 3.0349, 1.4712, 2.9174, 1.2250, 3.2048], + device='cuda:1'), covar=tensor([0.0061, 0.0971, 0.1814, 0.0025, 0.4692, 0.0047, 0.1313, 0.0078], + device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0269, 0.0321, 0.0132, 0.0314, 0.0145, 0.0275, 0.0175], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 20:41:40,765 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23041.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:41:41,721 INFO [train.py:901] (1/2) Epoch 9, batch 450, loss[loss=0.1919, simple_loss=0.2617, pruned_loss=0.06098, over 7314.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.262, pruned_loss=0.06578, over 1293708.51 frames. ], batch size: 49, lr: 1.78e-02, grad_scale: 16.0 +2023-03-20 20:41:45,849 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0533, 3.9600, 3.4470, 3.9004, 2.9084, 2.7343, 3.9497, 3.1522], + device='cuda:1'), covar=tensor([0.0111, 0.0110, 0.0167, 0.0089, 0.0309, 0.0355, 0.0144, 0.0517], + device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0230, 0.0229, 0.0227, 0.0284, 0.0289, 0.0245, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 20:41:49,280 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 20:41:50,369 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 20:41:53,912 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6979, 3.7777, 3.7324, 3.6747, 3.6064, 3.6630, 3.8265, 3.9982], + device='cuda:1'), covar=tensor([0.0356, 0.0286, 0.0385, 0.0479, 0.0555, 0.0456, 0.0491, 0.0407], + device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0095, 0.0090, 0.0100, 0.0097, 0.0078, 0.0077, 0.0074], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 20:42:00,743 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 20:42:01,240 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.20 vs. limit=5.0 +2023-03-20 20:42:02,543 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0259, 2.0012, 1.9104, 3.0344, 1.5662, 2.8544, 1.2110, 3.1703], + device='cuda:1'), covar=tensor([0.0042, 0.0874, 0.2064, 0.0031, 0.4141, 0.0041, 0.1160, 0.0088], + device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0267, 0.0319, 0.0132, 0.0311, 0.0145, 0.0274, 0.0175], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 20:42:04,759 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 +2023-03-20 20:42:05,461 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=23089.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:42:07,488 INFO [train.py:901] (1/2) Epoch 9, batch 500, loss[loss=0.1327, simple_loss=0.1886, pruned_loss=0.03836, over 6123.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2626, pruned_loss=0.06574, over 1328292.34 frames. ], batch size: 26, lr: 1.78e-02, grad_scale: 16.0 +2023-03-20 20:42:12,206 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6375, 2.6250, 3.0732, 2.9595, 2.9278, 2.8251, 2.1893, 2.9989], + device='cuda:1'), covar=tensor([0.2072, 0.0630, 0.1095, 0.1769, 0.1092, 0.1722, 0.3815, 0.1373], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0029, 0.0030, 0.0030, 0.0026, 0.0027, 0.0038, 0.0027], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 20:42:15,279 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23108.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:42:15,616 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.581e+02 2.594e+02 3.365e+02 3.863e+02 9.016e+02, threshold=6.731e+02, percent-clipped=1.0 +2023-03-20 20:42:22,209 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 20:42:23,715 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 20:42:24,219 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 20:42:26,790 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 20:42:31,279 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 20:42:33,266 INFO [train.py:901] (1/2) Epoch 9, batch 550, loss[loss=0.1873, simple_loss=0.2546, pruned_loss=0.05997, over 7222.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2615, pruned_loss=0.06534, over 1352983.67 frames. ], batch size: 45, lr: 1.78e-02, grad_scale: 16.0 +2023-03-20 20:42:37,292 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0094, 1.0094, 0.9193, 1.0626, 1.1138, 0.8688, 0.7971, 0.7615], + device='cuda:1'), covar=tensor([0.0182, 0.0093, 0.0179, 0.0080, 0.0151, 0.0108, 0.0130, 0.0204], + device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0017, 0.0016, 0.0019, 0.0021, 0.0018, 0.0018, 0.0021], + device='cuda:1'), out_proj_covar=tensor([2.5160e-05, 2.0540e-05, 2.1202e-05, 2.1260e-05, 2.4566e-05, 2.0754e-05, + 2.1767e-05, 2.8656e-05], device='cuda:1') +2023-03-20 20:42:37,721 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23150.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:42:43,259 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 20:42:47,406 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23169.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:42:51,216 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 20:42:53,223 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23181.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:42:54,700 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 20:42:55,855 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.9785, 0.8912, 0.9873, 1.2740, 0.9127, 1.1485, 0.8697, 1.0283], + device='cuda:1'), covar=tensor([0.1343, 0.2137, 0.0754, 0.0573, 0.1656, 0.1120, 0.0779, 0.1081], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0039, 0.0032, 0.0031, 0.0030, 0.0031, 0.0038, 0.0035], + device='cuda:1'), out_proj_covar=tensor([6.3992e-05, 8.5147e-05, 6.1396e-05, 6.1231e-05, 6.4176e-05, 6.5096e-05, + 7.6547e-05, 7.5000e-05], device='cuda:1') +2023-03-20 20:42:59,277 INFO [train.py:901] (1/2) Epoch 9, batch 600, loss[loss=0.1957, simple_loss=0.268, pruned_loss=0.06171, over 7274.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2624, pruned_loss=0.06533, over 1374160.65 frames. ], batch size: 52, lr: 1.78e-02, grad_scale: 16.0 +2023-03-20 20:43:01,814 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 20:43:07,643 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.756e+02 2.964e+02 3.775e+02 5.004e+02 1.113e+03, threshold=7.551e+02, percent-clipped=10.0 +2023-03-20 20:43:12,408 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23217.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:43:18,382 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23229.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:43:19,260 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 20:43:25,891 INFO [train.py:901] (1/2) Epoch 9, batch 650, loss[loss=0.1887, simple_loss=0.2573, pruned_loss=0.06008, over 7325.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2615, pruned_loss=0.06543, over 1387646.98 frames. ], batch size: 54, lr: 1.78e-02, grad_scale: 16.0 +2023-03-20 20:43:28,973 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 20:43:43,117 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6785, 4.2940, 4.2298, 3.8464, 4.2730, 2.8043, 2.1408, 4.5246], + device='cuda:1'), covar=tensor([0.0008, 0.0104, 0.0039, 0.0048, 0.0019, 0.0348, 0.0459, 0.0027], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0056, 0.0080, 0.0064, 0.0075, 0.0099, 0.0107, 0.0067], + device='cuda:1'), out_proj_covar=tensor([7.2710e-05, 8.7308e-05, 1.1661e-04, 9.5922e-05, 1.0423e-04, 1.4372e-04, + 1.5344e-04, 9.5116e-05], device='cuda:1') +2023-03-20 20:43:43,632 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23278.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:43:45,958 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 20:43:49,540 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23290.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:43:50,898 INFO [train.py:901] (1/2) Epoch 9, batch 700, loss[loss=0.2077, simple_loss=0.2653, pruned_loss=0.07508, over 7242.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2621, pruned_loss=0.06558, over 1399191.00 frames. ], batch size: 45, lr: 1.77e-02, grad_scale: 16.0 +2023-03-20 20:43:51,992 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23295.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:43:53,938 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 20:43:59,404 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.653e+02 2.801e+02 3.274e+02 4.080e+02 8.180e+02, threshold=6.549e+02, percent-clipped=2.0 +2023-03-20 20:44:17,138 INFO [train.py:901] (1/2) Epoch 9, batch 750, loss[loss=0.1636, simple_loss=0.2219, pruned_loss=0.0527, over 7015.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2621, pruned_loss=0.06551, over 1408027.67 frames. ], batch size: 35, lr: 1.77e-02, grad_scale: 16.0 +2023-03-20 20:44:17,192 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=23343.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:44:18,159 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 20:44:18,663 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 20:44:31,737 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 20:44:35,828 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 20:44:42,397 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 20:44:42,841 INFO [train.py:901] (1/2) Epoch 9, batch 800, loss[loss=0.2205, simple_loss=0.2842, pruned_loss=0.07839, over 7253.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2623, pruned_loss=0.06555, over 1416059.88 frames. ], batch size: 64, lr: 1.77e-02, grad_scale: 16.0 +2023-03-20 20:44:43,340 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 20:44:51,714 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.882e+02 3.186e+02 4.037e+02 8.134e+02, threshold=6.373e+02, percent-clipped=1.0 +2023-03-20 20:44:54,201 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 20:45:03,329 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1618, 1.1172, 1.0186, 1.6039, 1.3244, 1.3758, 1.0964, 1.2099], + device='cuda:1'), covar=tensor([0.0960, 0.1589, 0.0844, 0.0585, 0.0909, 0.0695, 0.0808, 0.1386], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0037, 0.0030, 0.0030, 0.0027, 0.0029, 0.0037, 0.0034], + device='cuda:1'), out_proj_covar=tensor([6.0375e-05, 8.1233e-05, 5.8705e-05, 5.8572e-05, 5.9789e-05, 6.1970e-05, + 7.5209e-05, 7.2778e-05], device='cuda:1') +2023-03-20 20:45:08,671 INFO [train.py:901] (1/2) Epoch 9, batch 850, loss[loss=0.2058, simple_loss=0.2703, pruned_loss=0.07067, over 7268.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2632, pruned_loss=0.06603, over 1422784.14 frames. ], batch size: 47, lr: 1.77e-02, grad_scale: 16.0 +2023-03-20 20:45:12,195 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23450.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:45:12,582 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 20:45:12,592 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 20:45:17,764 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 20:45:19,330 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23464.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:45:21,330 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 20:45:28,452 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23481.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:45:32,610 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23488.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:45:34,968 INFO [train.py:901] (1/2) Epoch 9, batch 900, loss[loss=0.1967, simple_loss=0.2651, pruned_loss=0.06411, over 7281.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2627, pruned_loss=0.06565, over 1426879.89 frames. ], batch size: 52, lr: 1.77e-02, grad_scale: 16.0 +2023-03-20 20:45:37,495 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=23498.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:45:42,904 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.727e+02 2.905e+02 3.478e+02 4.319e+02 1.982e+03, threshold=6.955e+02, percent-clipped=5.0 +2023-03-20 20:45:52,949 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=23529.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:45:58,143 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9789, 3.9530, 3.5011, 3.2865, 3.5679, 2.2664, 1.5383, 3.8797], + device='cuda:1'), covar=tensor([0.0013, 0.0038, 0.0053, 0.0064, 0.0039, 0.0376, 0.0606, 0.0034], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0054, 0.0080, 0.0064, 0.0075, 0.0099, 0.0105, 0.0066], + device='cuda:1'), out_proj_covar=tensor([7.2184e-05, 8.4129e-05, 1.1493e-04, 9.5025e-05, 1.0246e-04, 1.4271e-04, + 1.5097e-04, 9.4595e-05], device='cuda:1') +2023-03-20 20:45:59,554 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 20:46:00,027 INFO [train.py:901] (1/2) Epoch 9, batch 950, loss[loss=0.1917, simple_loss=0.2731, pruned_loss=0.05513, over 7138.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2622, pruned_loss=0.06518, over 1428455.26 frames. ], batch size: 98, lr: 1.77e-02, grad_scale: 16.0 +2023-03-20 20:46:03,143 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23549.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 20:46:16,349 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23573.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:46:22,312 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23585.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:46:23,723 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 20:46:26,207 INFO [train.py:901] (1/2) Epoch 9, batch 1000, loss[loss=0.2236, simple_loss=0.2845, pruned_loss=0.08134, over 6641.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2624, pruned_loss=0.06569, over 1432917.28 frames. ], batch size: 106, lr: 1.76e-02, grad_scale: 16.0 +2023-03-20 20:46:34,369 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.714e+02 2.956e+02 3.607e+02 4.616e+02 9.211e+02, threshold=7.213e+02, percent-clipped=2.0 +2023-03-20 20:46:43,830 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 20:46:51,212 INFO [train.py:901] (1/2) Epoch 9, batch 1050, loss[loss=0.2089, simple_loss=0.2854, pruned_loss=0.06624, over 7298.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2628, pruned_loss=0.0659, over 1434587.67 frames. ], batch size: 80, lr: 1.76e-02, grad_scale: 16.0 +2023-03-20 20:47:05,631 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 20:47:09,646 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 20:47:10,295 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1509, 0.9507, 1.1634, 1.3350, 1.2058, 1.3230, 0.9648, 1.0891], + device='cuda:1'), covar=tensor([0.0679, 0.1043, 0.0361, 0.0400, 0.0811, 0.0873, 0.0342, 0.1221], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0039, 0.0030, 0.0030, 0.0029, 0.0030, 0.0038, 0.0035], + device='cuda:1'), out_proj_covar=tensor([6.1821e-05, 8.3171e-05, 5.8537e-05, 6.0202e-05, 6.2575e-05, 6.3458e-05, + 7.6142e-05, 7.4176e-05], device='cuda:1') +2023-03-20 20:47:10,323 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7761, 2.2511, 1.9789, 3.2132, 2.9882, 3.0000, 2.7165, 2.9976], + device='cuda:1'), covar=tensor([0.1251, 0.0509, 0.1418, 0.0240, 0.0039, 0.0036, 0.0036, 0.0051], + device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0218, 0.0269, 0.0234, 0.0122, 0.0119, 0.0121, 0.0136], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 20:47:17,728 INFO [train.py:901] (1/2) Epoch 9, batch 1100, loss[loss=0.2258, simple_loss=0.2837, pruned_loss=0.08398, over 7250.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2619, pruned_loss=0.06536, over 1436024.19 frames. ], batch size: 55, lr: 1.76e-02, grad_scale: 16.0 +2023-03-20 20:47:25,695 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.755e+02 2.952e+02 3.526e+02 4.290e+02 1.100e+03, threshold=7.053e+02, percent-clipped=3.0 +2023-03-20 20:47:28,408 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.9751, 1.1796, 1.2342, 1.1788, 1.2181, 0.8055, 0.9571, 0.7775], + device='cuda:1'), covar=tensor([0.0129, 0.0173, 0.0125, 0.0082, 0.0139, 0.0087, 0.0150, 0.0239], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0019, 0.0019, 0.0019, 0.0021, 0.0019, 0.0020, 0.0023], + device='cuda:1'), out_proj_covar=tensor([2.6153e-05, 2.2204e-05, 2.3519e-05, 2.2021e-05, 2.5441e-05, 2.1901e-05, + 2.2928e-05, 3.0822e-05], device='cuda:1') +2023-03-20 20:47:38,331 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 20:47:38,873 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 20:47:43,506 INFO [train.py:901] (1/2) Epoch 9, batch 1150, loss[loss=0.1561, simple_loss=0.2326, pruned_loss=0.03977, over 7169.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2626, pruned_loss=0.06581, over 1437017.23 frames. ], batch size: 39, lr: 1.76e-02, grad_scale: 16.0 +2023-03-20 20:47:52,130 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 20:47:52,685 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 20:47:54,738 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23764.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:47:57,684 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4001, 4.8690, 4.8795, 4.7669, 4.6040, 4.4983, 4.8791, 4.6840], + device='cuda:1'), covar=tensor([0.0386, 0.0312, 0.0299, 0.0373, 0.0311, 0.0270, 0.0295, 0.0445], + device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0151, 0.0115, 0.0114, 0.0100, 0.0145, 0.0128, 0.0106], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 20:47:59,372 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-20 20:48:04,806 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4411, 1.7257, 1.7953, 1.2868, 1.0124, 1.7167, 1.5746, 1.2769], + device='cuda:1'), covar=tensor([0.0297, 0.0315, 0.0171, 0.0110, 0.0791, 0.0379, 0.0118, 0.0291], + device='cuda:1'), in_proj_covar=tensor([0.0017, 0.0018, 0.0017, 0.0017, 0.0018, 0.0016, 0.0017, 0.0017], + device='cuda:1'), out_proj_covar=tensor([4.1878e-05, 4.1825e-05, 3.9047e-05, 3.5896e-05, 4.2645e-05, 3.7160e-05, + 3.9614e-05, 4.2967e-05], device='cuda:1') +2023-03-20 20:48:09,158 INFO [train.py:901] (1/2) Epoch 9, batch 1200, loss[loss=0.2079, simple_loss=0.2678, pruned_loss=0.07398, over 7258.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2636, pruned_loss=0.06645, over 1439815.75 frames. ], batch size: 89, lr: 1.76e-02, grad_scale: 16.0 +2023-03-20 20:48:17,480 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.614e+02 2.908e+02 3.769e+02 4.878e+02 9.348e+02, threshold=7.537e+02, percent-clipped=4.0 +2023-03-20 20:48:19,052 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=23812.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:48:21,646 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0681, 4.5253, 4.5518, 4.4543, 4.4283, 4.1440, 4.5335, 4.3995], + device='cuda:1'), covar=tensor([0.0423, 0.0440, 0.0402, 0.0439, 0.0388, 0.0332, 0.0412, 0.0589], + device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0149, 0.0114, 0.0114, 0.0100, 0.0142, 0.0125, 0.0105], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 20:48:24,157 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 20:48:35,830 INFO [train.py:901] (1/2) Epoch 9, batch 1250, loss[loss=0.209, simple_loss=0.2725, pruned_loss=0.0727, over 7278.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2625, pruned_loss=0.06545, over 1439215.28 frames. ], batch size: 77, lr: 1.75e-02, grad_scale: 16.0 +2023-03-20 20:48:36,426 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23844.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 20:48:36,886 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0616, 3.9440, 3.9694, 4.4190, 4.3324, 4.3409, 3.6448, 3.8189], + device='cuda:1'), covar=tensor([0.1142, 0.2599, 0.2149, 0.0949, 0.0724, 0.1443, 0.0794, 0.1102], + device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0257, 0.0222, 0.0202, 0.0158, 0.0271, 0.0148, 0.0180], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002], + device='cuda:1') +2023-03-20 20:48:48,742 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 20:48:50,906 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23873.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:48:53,331 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 20:48:54,812 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 20:48:56,974 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23885.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:49:00,501 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.9120, 0.9572, 1.1195, 1.4574, 1.2170, 1.1806, 0.9878, 1.1450], + device='cuda:1'), covar=tensor([0.1178, 0.1988, 0.0915, 0.0703, 0.1508, 0.1241, 0.0962, 0.2034], + device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0038, 0.0031, 0.0031, 0.0030, 0.0030, 0.0038, 0.0034], + device='cuda:1'), out_proj_covar=tensor([6.2740e-05, 8.3513e-05, 6.1173e-05, 6.0447e-05, 6.5086e-05, 6.4797e-05, + 7.8842e-05, 7.3062e-05], device='cuda:1') +2023-03-20 20:49:00,813 INFO [train.py:901] (1/2) Epoch 9, batch 1300, loss[loss=0.2136, simple_loss=0.2709, pruned_loss=0.07815, over 7315.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2619, pruned_loss=0.06534, over 1439111.42 frames. ], batch size: 49, lr: 1.75e-02, grad_scale: 16.0 +2023-03-20 20:49:09,404 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.579e+02 3.213e+02 4.106e+02 9.260e+02, threshold=6.425e+02, percent-clipped=2.0 +2023-03-20 20:49:15,605 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=23921.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:49:17,521 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 20:49:20,755 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 20:49:22,280 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=23933.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:49:24,276 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 20:49:27,207 INFO [train.py:901] (1/2) Epoch 9, batch 1350, loss[loss=0.2342, simple_loss=0.2928, pruned_loss=0.08787, over 7244.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2625, pruned_loss=0.06557, over 1440140.24 frames. ], batch size: 93, lr: 1.75e-02, grad_scale: 32.0 +2023-03-20 20:49:34,679 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 20:49:52,680 INFO [train.py:901] (1/2) Epoch 9, batch 1400, loss[loss=0.165, simple_loss=0.2431, pruned_loss=0.0434, over 7294.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2635, pruned_loss=0.06592, over 1438756.83 frames. ], batch size: 86, lr: 1.75e-02, grad_scale: 32.0 +2023-03-20 20:49:53,840 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0553, 1.2216, 1.3682, 1.3615, 1.4349, 0.8974, 0.6516, 0.8178], + device='cuda:1'), covar=tensor([0.0148, 0.0170, 0.0338, 0.0102, 0.0199, 0.0154, 0.0161, 0.0260], + device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0018, 0.0018, 0.0019, 0.0022, 0.0019, 0.0020, 0.0023], + device='cuda:1'), out_proj_covar=tensor([2.5535e-05, 2.1305e-05, 2.3035e-05, 2.2354e-05, 2.5622e-05, 2.2304e-05, + 2.2915e-05, 3.0886e-05], device='cuda:1') +2023-03-20 20:50:04,557 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.640e+02 2.794e+02 3.299e+02 4.063e+02 9.313e+02, threshold=6.598e+02, percent-clipped=3.0 +2023-03-20 20:50:10,221 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 20:50:19,383 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24037.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:50:20,477 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.8974, 1.0070, 0.9865, 1.2487, 1.2703, 1.2718, 0.8834, 0.9592], + device='cuda:1'), covar=tensor([0.0715, 0.1734, 0.0596, 0.0535, 0.0675, 0.0609, 0.0869, 0.1300], + device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0038, 0.0031, 0.0031, 0.0030, 0.0031, 0.0038, 0.0034], + device='cuda:1'), out_proj_covar=tensor([6.2702e-05, 8.3092e-05, 6.1817e-05, 6.0683e-05, 6.4567e-05, 6.5434e-05, + 7.8683e-05, 7.3269e-05], device='cuda:1') +2023-03-20 20:50:22,309 INFO [train.py:901] (1/2) Epoch 9, batch 1450, loss[loss=0.1687, simple_loss=0.2453, pruned_loss=0.04605, over 7349.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2615, pruned_loss=0.06487, over 1437209.59 frames. ], batch size: 73, lr: 1.75e-02, grad_scale: 32.0 +2023-03-20 20:50:33,916 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 20:50:47,918 INFO [train.py:901] (1/2) Epoch 9, batch 1500, loss[loss=0.2113, simple_loss=0.272, pruned_loss=0.07532, over 7308.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2619, pruned_loss=0.06502, over 1440527.66 frames. ], batch size: 83, lr: 1.75e-02, grad_scale: 32.0 +2023-03-20 20:50:50,390 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 20:50:50,542 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24098.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:50:56,467 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.802e+02 2.774e+02 3.566e+02 4.390e+02 1.132e+03, threshold=7.132e+02, percent-clipped=5.0 +2023-03-20 20:51:01,679 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8138, 2.5094, 2.6980, 2.8017, 3.0809, 2.4089, 1.9496, 2.8869], + device='cuda:1'), covar=tensor([0.1320, 0.0488, 0.1342, 0.1814, 0.0907, 0.2239, 0.3495, 0.1470], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0030, 0.0031, 0.0031, 0.0027, 0.0029, 0.0041, 0.0029], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 20:51:05,636 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8379, 2.4789, 2.7793, 2.8277, 3.0336, 2.3561, 1.9780, 2.8623], + device='cuda:1'), covar=tensor([0.1183, 0.0436, 0.1392, 0.1658, 0.0858, 0.2383, 0.3445, 0.1509], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0030, 0.0031, 0.0031, 0.0027, 0.0029, 0.0041, 0.0029], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 20:51:13,480 INFO [train.py:901] (1/2) Epoch 9, batch 1550, loss[loss=0.1847, simple_loss=0.2399, pruned_loss=0.06474, over 7198.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2622, pruned_loss=0.06483, over 1441293.25 frames. ], batch size: 39, lr: 1.74e-02, grad_scale: 32.0 +2023-03-20 20:51:14,123 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24144.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 20:51:15,014 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 20:51:39,445 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=24192.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:51:39,892 INFO [train.py:901] (1/2) Epoch 9, batch 1600, loss[loss=0.1689, simple_loss=0.2362, pruned_loss=0.05079, over 7285.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.262, pruned_loss=0.06465, over 1442950.89 frames. ], batch size: 47, lr: 1.74e-02, grad_scale: 32.0 +2023-03-20 20:51:45,879 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1428, 4.5825, 4.5984, 4.4909, 4.4612, 4.1479, 4.5992, 4.4378], + device='cuda:1'), covar=tensor([0.0433, 0.0414, 0.0369, 0.0464, 0.0308, 0.0367, 0.0424, 0.0515], + device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0157, 0.0119, 0.0119, 0.0104, 0.0150, 0.0133, 0.0110], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 20:51:46,821 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 20:51:47,843 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 20:51:48,304 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.692e+02 2.929e+02 3.453e+02 4.111e+02 6.914e+02, threshold=6.906e+02, percent-clipped=0.0 +2023-03-20 20:51:50,864 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 20:51:59,375 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 20:52:03,413 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 20:52:05,439 INFO [train.py:901] (1/2) Epoch 9, batch 1650, loss[loss=0.1643, simple_loss=0.2359, pruned_loss=0.04634, over 7314.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2617, pruned_loss=0.06417, over 1444560.03 frames. ], batch size: 44, lr: 1.74e-02, grad_scale: 32.0 +2023-03-20 20:52:12,836 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 20:52:14,779 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 +2023-03-20 20:52:29,834 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-20 20:52:30,083 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 20:52:32,056 INFO [train.py:901] (1/2) Epoch 9, batch 1700, loss[loss=0.1625, simple_loss=0.2372, pruned_loss=0.04386, over 7309.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2606, pruned_loss=0.06381, over 1439524.98 frames. ], batch size: 42, lr: 1.74e-02, grad_scale: 32.0 +2023-03-20 20:52:34,096 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 20:52:40,052 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.835e+02 3.245e+02 4.182e+02 9.233e+02, threshold=6.490e+02, percent-clipped=3.0 +2023-03-20 20:52:44,519 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 20:52:57,650 INFO [train.py:901] (1/2) Epoch 9, batch 1750, loss[loss=0.2207, simple_loss=0.2897, pruned_loss=0.07583, over 7238.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.26, pruned_loss=0.06344, over 1441444.35 frames. ], batch size: 93, lr: 1.74e-02, grad_scale: 32.0 +2023-03-20 20:53:08,788 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 20:53:10,422 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 20:53:18,013 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4553, 3.0187, 3.1374, 3.4069, 3.3548, 3.3505, 3.0552, 3.0580], + device='cuda:1'), covar=tensor([0.0045, 0.0133, 0.0076, 0.0056, 0.0056, 0.0073, 0.0139, 0.0125], + device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0037, 0.0035, 0.0033, 0.0035, 0.0035, 0.0045, 0.0042], + device='cuda:1'), out_proj_covar=tensor([8.2589e-05, 1.1364e-04, 1.1196e-04, 9.3380e-05, 1.0219e-04, 1.0088e-04, + 1.4290e-04, 1.2289e-04], device='cuda:1') +2023-03-20 20:53:23,360 INFO [train.py:901] (1/2) Epoch 9, batch 1800, loss[loss=0.2006, simple_loss=0.2743, pruned_loss=0.06349, over 7258.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2604, pruned_loss=0.06381, over 1441097.08 frames. ], batch size: 89, lr: 1.74e-02, grad_scale: 32.0 +2023-03-20 20:53:23,444 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24393.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:53:31,260 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 20:53:31,745 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.774e+02 2.625e+02 3.149e+02 3.948e+02 8.480e+02, threshold=6.298e+02, percent-clipped=3.0 +2023-03-20 20:53:44,885 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 20:53:49,368 INFO [train.py:901] (1/2) Epoch 9, batch 1850, loss[loss=0.152, simple_loss=0.2094, pruned_loss=0.04733, over 6159.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2605, pruned_loss=0.06404, over 1439551.50 frames. ], batch size: 26, lr: 1.73e-02, grad_scale: 16.0 +2023-03-20 20:53:56,089 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 20:54:02,924 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.04 vs. limit=2.0 +2023-03-20 20:54:12,099 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 20:54:13,327 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9844, 2.2798, 2.0367, 3.0948, 2.4615, 2.8747, 2.5018, 2.9347], + device='cuda:1'), covar=tensor([0.1410, 0.0672, 0.1863, 0.0217, 0.0056, 0.0044, 0.0049, 0.0055], + device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0220, 0.0274, 0.0235, 0.0120, 0.0120, 0.0125, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 20:54:15,087 INFO [train.py:901] (1/2) Epoch 9, batch 1900, loss[loss=0.1948, simple_loss=0.267, pruned_loss=0.06124, over 7225.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2603, pruned_loss=0.06388, over 1440633.29 frames. ], batch size: 93, lr: 1.73e-02, grad_scale: 16.0 +2023-03-20 20:54:23,574 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.695e+02 2.705e+02 3.277e+02 3.704e+02 6.252e+02, threshold=6.553e+02, percent-clipped=0.0 +2023-03-20 20:54:37,818 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 20:54:41,449 INFO [train.py:901] (1/2) Epoch 9, batch 1950, loss[loss=0.1531, simple_loss=0.1992, pruned_loss=0.05349, over 6429.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2594, pruned_loss=0.06325, over 1439728.63 frames. ], batch size: 28, lr: 1.73e-02, grad_scale: 16.0 +2023-03-20 20:54:43,616 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24547.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:54:49,119 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 20:54:53,707 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 20:54:54,188 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 20:54:54,807 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24569.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:55:05,468 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6571, 2.1709, 2.2102, 3.3202, 1.4737, 3.3017, 1.3413, 3.3643], + device='cuda:1'), covar=tensor([0.0048, 0.0831, 0.1733, 0.0035, 0.4158, 0.0048, 0.1042, 0.0084], + device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0258, 0.0308, 0.0136, 0.0296, 0.0138, 0.0267, 0.0170], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0001, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 20:55:06,810 INFO [train.py:901] (1/2) Epoch 9, batch 2000, loss[loss=0.1961, simple_loss=0.2567, pruned_loss=0.06776, over 7310.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2601, pruned_loss=0.06376, over 1441340.76 frames. ], batch size: 83, lr: 1.73e-02, grad_scale: 16.0 +2023-03-20 20:55:10,344 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 20:55:15,447 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24608.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:55:16,212 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.744e+02 2.663e+02 3.037e+02 3.690e+02 7.050e+02, threshold=6.074e+02, percent-clipped=1.0 +2023-03-20 20:55:22,191 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 20:55:27,103 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24630.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:55:27,608 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8082, 2.1680, 1.9753, 3.2157, 2.7753, 2.9304, 2.6008, 3.0519], + device='cuda:1'), covar=tensor([0.1786, 0.0761, 0.2072, 0.0291, 0.0073, 0.0088, 0.0070, 0.0122], + device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0226, 0.0279, 0.0240, 0.0122, 0.0124, 0.0130, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-03-20 20:55:30,391 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 20:55:33,471 INFO [train.py:901] (1/2) Epoch 9, batch 2050, loss[loss=0.2076, simple_loss=0.2726, pruned_loss=0.07127, over 7290.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2608, pruned_loss=0.06405, over 1441653.65 frames. ], batch size: 66, lr: 1.73e-02, grad_scale: 16.0 +2023-03-20 20:55:49,535 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1785, 0.9718, 1.0817, 1.1932, 1.1861, 1.3662, 1.0808, 1.0724], + device='cuda:1'), covar=tensor([0.0822, 0.0925, 0.0948, 0.0804, 0.0828, 0.1245, 0.0591, 0.1300], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0036, 0.0029, 0.0029, 0.0029, 0.0028, 0.0036, 0.0032], + device='cuda:1'), out_proj_covar=tensor([6.0432e-05, 7.9168e-05, 5.9404e-05, 5.9598e-05, 6.2677e-05, 6.2294e-05, + 7.6056e-05, 6.9764e-05], device='cuda:1') +2023-03-20 20:55:58,935 INFO [train.py:901] (1/2) Epoch 9, batch 2100, loss[loss=0.1924, simple_loss=0.2616, pruned_loss=0.06156, over 7329.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2607, pruned_loss=0.06408, over 1443187.71 frames. ], batch size: 54, lr: 1.73e-02, grad_scale: 16.0 +2023-03-20 20:55:59,051 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24693.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:56:03,932 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 20:56:07,404 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.780e+02 2.739e+02 3.522e+02 4.404e+02 7.179e+02, threshold=7.043e+02, percent-clipped=4.0 +2023-03-20 20:56:07,425 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 20:56:11,729 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2658, 3.7923, 3.8214, 3.8148, 3.6201, 3.9436, 4.1605, 3.5416], + device='cuda:1'), covar=tensor([0.0097, 0.0105, 0.0152, 0.0123, 0.0218, 0.0076, 0.0124, 0.0133], + device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0053, 0.0056, 0.0044, 0.0077, 0.0059, 0.0056, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 20:56:23,703 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=24741.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:56:24,615 INFO [train.py:901] (1/2) Epoch 9, batch 2150, loss[loss=0.2057, simple_loss=0.2659, pruned_loss=0.07279, over 7259.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2612, pruned_loss=0.06425, over 1444109.62 frames. ], batch size: 77, lr: 1.72e-02, grad_scale: 16.0 +2023-03-20 20:56:35,918 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24765.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:56:40,959 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3489, 1.7048, 1.3193, 1.3280, 1.5224, 1.0513, 0.9155, 0.8815], + device='cuda:1'), covar=tensor([0.0211, 0.0156, 0.0203, 0.0146, 0.0098, 0.0234, 0.0225, 0.0261], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0018, 0.0018, 0.0019, 0.0020, 0.0019, 0.0019, 0.0023], + device='cuda:1'), out_proj_covar=tensor([2.6447e-05, 2.0818e-05, 2.3159e-05, 2.2088e-05, 2.4327e-05, 2.2036e-05, + 2.1616e-05, 3.0273e-05], device='cuda:1') +2023-03-20 20:56:50,678 INFO [train.py:901] (1/2) Epoch 9, batch 2200, loss[loss=0.1931, simple_loss=0.2632, pruned_loss=0.06155, over 7333.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2608, pruned_loss=0.06408, over 1442199.63 frames. ], batch size: 59, lr: 1.72e-02, grad_scale: 16.0 +2023-03-20 20:56:52,302 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0703, 2.2603, 1.7059, 1.7156, 2.0778, 2.0491, 2.2198, 2.0357], + device='cuda:1'), covar=tensor([0.3504, 0.1137, 0.1922, 0.1828, 0.1696, 0.1035, 0.1392, 0.1893], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0033, 0.0040, 0.0037, 0.0039, 0.0036, 0.0035, 0.0035], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 20:56:52,689 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 20:57:00,150 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.794e+02 2.770e+02 3.163e+02 3.850e+02 5.609e+02, threshold=6.326e+02, percent-clipped=0.0 +2023-03-20 20:57:08,502 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24826.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:57:16,955 INFO [train.py:901] (1/2) Epoch 9, batch 2250, loss[loss=0.1922, simple_loss=0.2622, pruned_loss=0.06104, over 7246.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2603, pruned_loss=0.06359, over 1441715.28 frames. ], batch size: 89, lr: 1.72e-02, grad_scale: 16.0 +2023-03-20 20:57:19,754 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24848.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:57:27,768 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 20:57:27,779 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 20:57:38,482 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9121, 2.0807, 1.8347, 2.9526, 2.4938, 2.4947, 2.7272, 2.6600], + device='cuda:1'), covar=tensor([0.1322, 0.0601, 0.1946, 0.0338, 0.0042, 0.0030, 0.0041, 0.0051], + device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0226, 0.0281, 0.0239, 0.0122, 0.0121, 0.0133, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-03-20 20:57:40,459 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 20:57:41,142 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3827, 4.2896, 3.8561, 4.1511, 3.1462, 3.0717, 4.4102, 3.4596], + device='cuda:1'), covar=tensor([0.0112, 0.0111, 0.0103, 0.0094, 0.0313, 0.0389, 0.0189, 0.0399], + device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0248, 0.0231, 0.0241, 0.0291, 0.0286, 0.0253, 0.0295], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 20:57:43,439 INFO [train.py:901] (1/2) Epoch 9, batch 2300, loss[loss=0.219, simple_loss=0.2742, pruned_loss=0.0819, over 7273.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2596, pruned_loss=0.06326, over 1441236.88 frames. ], batch size: 57, lr: 1.72e-02, grad_scale: 16.0 +2023-03-20 20:57:48,453 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24903.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:57:51,549 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24909.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:57:51,864 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.909e+02 3.420e+02 4.609e+02 1.423e+03, threshold=6.841e+02, percent-clipped=7.0 +2023-03-20 20:57:59,528 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24925.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:58:08,493 INFO [train.py:901] (1/2) Epoch 9, batch 2350, loss[loss=0.1907, simple_loss=0.2667, pruned_loss=0.05734, over 7269.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2612, pruned_loss=0.06449, over 1441233.68 frames. ], batch size: 52, lr: 1.72e-02, grad_scale: 16.0 +2023-03-20 20:58:27,379 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 20:58:33,932 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 20:58:34,910 INFO [train.py:901] (1/2) Epoch 9, batch 2400, loss[loss=0.1805, simple_loss=0.2477, pruned_loss=0.05663, over 7281.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2608, pruned_loss=0.06385, over 1441322.83 frames. ], batch size: 70, lr: 1.72e-02, grad_scale: 16.0 +2023-03-20 20:58:36,025 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24995.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:58:41,072 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-03-20 20:58:43,807 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.523e+02 2.636e+02 3.211e+02 4.206e+02 8.943e+02, threshold=6.422e+02, percent-clipped=2.0 +2023-03-20 20:58:44,337 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 20:58:46,861 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 20:58:57,656 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9199, 2.3056, 1.8252, 3.2337, 2.5390, 2.8375, 2.7984, 2.9889], + device='cuda:1'), covar=tensor([0.1416, 0.0552, 0.2003, 0.0192, 0.0024, 0.0053, 0.0069, 0.0075], + device='cuda:1'), in_proj_covar=tensor([0.0298, 0.0225, 0.0280, 0.0236, 0.0120, 0.0122, 0.0135, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-03-20 20:59:01,012 INFO [train.py:901] (1/2) Epoch 9, batch 2450, loss[loss=0.1889, simple_loss=0.2576, pruned_loss=0.06007, over 7257.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2606, pruned_loss=0.06391, over 1440972.85 frames. ], batch size: 89, lr: 1.71e-02, grad_scale: 16.0 +2023-03-20 20:59:07,840 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25056.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 20:59:13,290 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 20:59:27,013 INFO [train.py:901] (1/2) Epoch 9, batch 2500, loss[loss=0.1912, simple_loss=0.2612, pruned_loss=0.06057, over 7325.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2608, pruned_loss=0.06384, over 1442547.10 frames. ], batch size: 59, lr: 1.71e-02, grad_scale: 16.0 +2023-03-20 20:59:35,682 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 2.803e+02 3.310e+02 3.977e+02 7.534e+02, threshold=6.620e+02, percent-clipped=2.0 +2023-03-20 20:59:37,699 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 20:59:41,580 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.07 vs. limit=5.0 +2023-03-20 20:59:41,876 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25121.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 20:59:52,969 INFO [train.py:901] (1/2) Epoch 9, batch 2550, loss[loss=0.165, simple_loss=0.224, pruned_loss=0.05302, over 6969.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.26, pruned_loss=0.06333, over 1440711.44 frames. ], batch size: 35, lr: 1.71e-02, grad_scale: 16.0 +2023-03-20 21:00:00,355 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-20 21:00:01,647 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1139, 4.5390, 4.5629, 4.3990, 4.4617, 4.0899, 4.5610, 4.4599], + device='cuda:1'), covar=tensor([0.0383, 0.0406, 0.0464, 0.0538, 0.0323, 0.0342, 0.0356, 0.0479], + device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0156, 0.0120, 0.0119, 0.0100, 0.0149, 0.0129, 0.0110], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:00:02,173 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7782, 3.3880, 3.3470, 3.6873, 3.6776, 3.9018, 3.4169, 3.4266], + device='cuda:1'), covar=tensor([0.0027, 0.0088, 0.0049, 0.0038, 0.0037, 0.0028, 0.0063, 0.0066], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0036, 0.0034, 0.0031, 0.0034, 0.0033, 0.0043, 0.0041], + device='cuda:1'), out_proj_covar=tensor([8.1409e-05, 1.1340e-04, 1.0736e-04, 8.7556e-05, 9.4820e-05, 9.3451e-05, + 1.3476e-04, 1.1951e-04], device='cuda:1') +2023-03-20 21:00:18,603 INFO [train.py:901] (1/2) Epoch 9, batch 2600, loss[loss=0.1936, simple_loss=0.26, pruned_loss=0.06361, over 7292.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2598, pruned_loss=0.06306, over 1439131.26 frames. ], batch size: 68, lr: 1.71e-02, grad_scale: 16.0 +2023-03-20 21:00:22,979 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1979, 2.1350, 2.1403, 3.3960, 1.3900, 3.3745, 1.3731, 3.2333], + device='cuda:1'), covar=tensor([0.0043, 0.0879, 0.1756, 0.0044, 0.4043, 0.0048, 0.1029, 0.0122], + device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0257, 0.0311, 0.0140, 0.0302, 0.0142, 0.0272, 0.0182], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 21:00:23,884 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25203.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:00:24,289 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25204.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:00:27,201 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.760e+02 2.724e+02 3.388e+02 4.447e+02 1.109e+03, threshold=6.776e+02, percent-clipped=3.0 +2023-03-20 21:00:29,915 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9154, 3.7328, 3.3149, 3.6196, 3.0337, 2.6376, 3.7389, 3.0368], + device='cuda:1'), covar=tensor([0.0126, 0.0114, 0.0136, 0.0109, 0.0224, 0.0302, 0.0193, 0.0367], + device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0244, 0.0224, 0.0237, 0.0289, 0.0283, 0.0255, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 21:00:34,845 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25225.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:00:43,538 INFO [train.py:901] (1/2) Epoch 9, batch 2650, loss[loss=0.1899, simple_loss=0.2556, pruned_loss=0.0621, over 7310.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.26, pruned_loss=0.06318, over 1440584.45 frames. ], batch size: 59, lr: 1.71e-02, grad_scale: 16.0 +2023-03-20 21:00:47,482 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=25251.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:00:58,865 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=25273.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:01:01,065 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 +2023-03-20 21:01:08,603 INFO [train.py:901] (1/2) Epoch 9, batch 2700, loss[loss=0.1443, simple_loss=0.2061, pruned_loss=0.04123, over 6998.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2594, pruned_loss=0.06322, over 1440085.88 frames. ], batch size: 35, lr: 1.71e-02, grad_scale: 16.0 +2023-03-20 21:01:16,824 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-03-20 21:01:17,022 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.798e+02 2.440e+02 3.297e+02 3.906e+02 8.338e+02, threshold=6.594e+02, percent-clipped=1.0 +2023-03-20 21:01:23,638 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9938, 3.9466, 3.5152, 3.8372, 3.0062, 2.6094, 4.0049, 3.1780], + device='cuda:1'), covar=tensor([0.0075, 0.0118, 0.0164, 0.0099, 0.0278, 0.0374, 0.0177, 0.0476], + device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0246, 0.0225, 0.0240, 0.0292, 0.0287, 0.0257, 0.0293], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 21:01:33,770 INFO [train.py:901] (1/2) Epoch 9, batch 2750, loss[loss=0.1609, simple_loss=0.2282, pruned_loss=0.04674, over 7170.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2595, pruned_loss=0.06337, over 1439751.65 frames. ], batch size: 39, lr: 1.70e-02, grad_scale: 16.0 +2023-03-20 21:01:37,757 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25351.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 21:01:57,911 INFO [train.py:901] (1/2) Epoch 9, batch 2800, loss[loss=0.2355, simple_loss=0.2986, pruned_loss=0.08616, over 7296.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2601, pruned_loss=0.06362, over 1442151.91 frames. ], batch size: 59, lr: 1.70e-02, grad_scale: 16.0 +2023-03-20 21:02:06,535 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 2.013e+02 2.834e+02 3.380e+02 4.238e+02 6.274e+02, threshold=6.760e+02, percent-clipped=0.0 +2023-03-20 21:02:23,093 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 21:02:23,147 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 21:02:23,170 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 21:02:23,188 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 21:02:23,359 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 21:02:23,374 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 21:02:23,385 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 21:02:23,482 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 21:02:23,902 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 21:02:23,917 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 21:02:23,937 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 21:02:24,094 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 21:02:24,099 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 21:02:24,107 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 21:02:24,244 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 21:02:24,246 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 21:02:24,304 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 21:02:31,135 INFO [train.py:901] (1/2) Epoch 10, batch 0, loss[loss=0.2043, simple_loss=0.2666, pruned_loss=0.07097, over 7346.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2666, pruned_loss=0.07097, over 7346.00 frames. ], batch size: 54, lr: 1.64e-02, grad_scale: 16.0 +2023-03-20 21:02:31,136 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 21:02:57,192 INFO [train.py:935] (1/2) Epoch 10, validation: loss=0.1762, simple_loss=0.2634, pruned_loss=0.04454, over 1622729.00 frames. +2023-03-20 21:02:57,193 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-20 21:02:59,262 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25421.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:03:02,784 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25428.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 21:03:03,673 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 21:03:14,207 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 21:03:18,823 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.9037, 1.7846, 2.3017, 1.6709, 1.2518, 1.4731, 1.4503, 1.1226], + device='cuda:1'), covar=tensor([0.0583, 0.0170, 0.0055, 0.0094, 0.0272, 0.0209, 0.0151, 0.0399], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0018, 0.0018, 0.0018, 0.0018, 0.0017, 0.0019, 0.0019], + device='cuda:1'), out_proj_covar=tensor([4.8880e-05, 4.3262e-05, 4.1271e-05, 3.7457e-05, 4.4187e-05, 4.0189e-05, + 4.3256e-05, 4.6737e-05], device='cuda:1') +2023-03-20 21:03:21,229 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 21:03:22,223 INFO [train.py:901] (1/2) Epoch 10, batch 50, loss[loss=0.1774, simple_loss=0.2509, pruned_loss=0.05199, over 7237.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2583, pruned_loss=0.06132, over 324247.28 frames. ], batch size: 93, lr: 1.63e-02, grad_scale: 16.0 +2023-03-20 21:03:23,237 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 21:03:23,285 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=25469.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:03:26,350 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 21:03:33,958 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25489.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 21:03:41,269 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25504.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:03:42,012 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.05 vs. limit=5.0 +2023-03-20 21:03:42,836 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 21:03:43,346 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 21:03:44,806 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.645e+02 3.259e+02 3.970e+02 7.214e+02, threshold=6.518e+02, percent-clipped=3.0 +2023-03-20 21:03:48,315 INFO [train.py:901] (1/2) Epoch 10, batch 100, loss[loss=0.2003, simple_loss=0.2692, pruned_loss=0.06571, over 7260.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2608, pruned_loss=0.06253, over 573507.97 frames. ], batch size: 64, lr: 1.63e-02, grad_scale: 16.0 +2023-03-20 21:03:59,510 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.7376, 0.7910, 1.0028, 1.1940, 0.9564, 1.1108, 0.9077, 1.0479], + device='cuda:1'), covar=tensor([0.0527, 0.1850, 0.1270, 0.0556, 0.1067, 0.2063, 0.0465, 0.2175], + device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0042, 0.0032, 0.0033, 0.0034, 0.0033, 0.0037, 0.0037], + device='cuda:1'), out_proj_covar=tensor([7.0221e-05, 9.1626e-05, 6.5830e-05, 6.7325e-05, 7.2898e-05, 7.2385e-05, + 8.1955e-05, 8.0184e-05], device='cuda:1') +2023-03-20 21:04:05,899 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=25552.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:04:13,854 INFO [train.py:901] (1/2) Epoch 10, batch 150, loss[loss=0.1663, simple_loss=0.2418, pruned_loss=0.04535, over 7308.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.261, pruned_loss=0.06317, over 767369.73 frames. ], batch size: 83, lr: 1.63e-02, grad_scale: 16.0 +2023-03-20 21:04:24,101 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2802, 1.4369, 1.3941, 1.2084, 1.4552, 0.9318, 0.9812, 0.9155], + device='cuda:1'), covar=tensor([0.0123, 0.0132, 0.0221, 0.0143, 0.0193, 0.0059, 0.0222, 0.0153], + device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0018, 0.0017, 0.0018, 0.0020, 0.0017, 0.0018, 0.0022], + device='cuda:1'), out_proj_covar=tensor([2.4836e-05, 2.0710e-05, 2.1807e-05, 2.0723e-05, 2.3789e-05, 1.9952e-05, + 2.1649e-05, 2.8321e-05], device='cuda:1') +2023-03-20 21:04:30,781 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 +2023-03-20 21:04:33,438 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9584, 3.9157, 3.4825, 3.4693, 3.4237, 2.2466, 1.8444, 3.9461], + device='cuda:1'), covar=tensor([0.0013, 0.0031, 0.0059, 0.0047, 0.0065, 0.0397, 0.0438, 0.0032], + device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0057, 0.0081, 0.0068, 0.0082, 0.0105, 0.0110, 0.0071], + device='cuda:1'), out_proj_covar=tensor([7.8193e-05, 8.8234e-05, 1.1547e-04, 1.0043e-04, 1.1193e-04, 1.5057e-04, + 1.5585e-04, 1.0133e-04], device='cuda:1') +2023-03-20 21:04:36,247 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.740e+02 3.249e+02 4.303e+02 1.148e+03, threshold=6.498e+02, percent-clipped=2.0 +2023-03-20 21:04:38,074 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.03 vs. limit=2.0 +2023-03-20 21:04:39,804 INFO [train.py:901] (1/2) Epoch 10, batch 200, loss[loss=0.1928, simple_loss=0.2685, pruned_loss=0.05855, over 7320.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2597, pruned_loss=0.06255, over 916314.32 frames. ], batch size: 83, lr: 1.63e-02, grad_scale: 16.0 +2023-03-20 21:04:40,136 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 +2023-03-20 21:04:41,834 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 21:04:46,790 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 21:04:52,896 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 21:04:57,377 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25651.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 21:05:05,340 INFO [train.py:901] (1/2) Epoch 10, batch 250, loss[loss=0.194, simple_loss=0.2625, pruned_loss=0.06275, over 7291.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2596, pruned_loss=0.06273, over 1030406.17 frames. ], batch size: 80, lr: 1.63e-02, grad_scale: 16.0 +2023-03-20 21:05:05,346 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 21:05:22,207 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=25699.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:05:26,631 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 21:05:27,635 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.691e+02 2.598e+02 3.271e+02 3.875e+02 1.164e+03, threshold=6.543e+02, percent-clipped=3.0 +2023-03-20 21:05:31,186 INFO [train.py:901] (1/2) Epoch 10, batch 300, loss[loss=0.1852, simple_loss=0.251, pruned_loss=0.05976, over 7262.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2589, pruned_loss=0.06272, over 1118531.93 frames. ], batch size: 47, lr: 1.63e-02, grad_scale: 16.0 +2023-03-20 21:05:35,731 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 21:05:57,625 INFO [train.py:901] (1/2) Epoch 10, batch 350, loss[loss=0.1839, simple_loss=0.25, pruned_loss=0.05889, over 7301.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2574, pruned_loss=0.06167, over 1190501.10 frames. ], batch size: 68, lr: 1.63e-02, grad_scale: 16.0 +2023-03-20 21:06:04,969 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 +2023-03-20 21:06:06,212 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25784.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 21:06:11,642 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 21:06:13,745 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0778, 1.7885, 2.6817, 1.8289, 1.2977, 1.4585, 1.8935, 1.4973], + device='cuda:1'), covar=tensor([0.0313, 0.0253, 0.0070, 0.0068, 0.0358, 0.0324, 0.0101, 0.0236], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0018, 0.0017, 0.0017, 0.0017, 0.0016, 0.0018, 0.0018], + device='cuda:1'), out_proj_covar=tensor([4.5577e-05, 4.3562e-05, 3.9743e-05, 3.5471e-05, 4.2403e-05, 3.8962e-05, + 4.1920e-05, 4.3871e-05], device='cuda:1') +2023-03-20 21:06:19,384 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.387e+02 2.765e+02 3.344e+02 4.046e+02 7.837e+02, threshold=6.687e+02, percent-clipped=1.0 +2023-03-20 21:06:23,524 INFO [train.py:901] (1/2) Epoch 10, batch 400, loss[loss=0.2164, simple_loss=0.2785, pruned_loss=0.07718, over 6750.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2576, pruned_loss=0.06182, over 1246270.76 frames. ], batch size: 107, lr: 1.62e-02, grad_scale: 16.0 +2023-03-20 21:06:44,819 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25858.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 21:06:49,125 INFO [train.py:901] (1/2) Epoch 10, batch 450, loss[loss=0.2032, simple_loss=0.269, pruned_loss=0.06869, over 7288.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2581, pruned_loss=0.0619, over 1291444.13 frames. ], batch size: 86, lr: 1.62e-02, grad_scale: 16.0 +2023-03-20 21:06:52,736 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 21:06:53,216 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 21:07:11,263 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.486e+02 2.544e+02 3.144e+02 3.954e+02 7.996e+02, threshold=6.287e+02, percent-clipped=4.0 +2023-03-20 21:07:14,821 INFO [train.py:901] (1/2) Epoch 10, batch 500, loss[loss=0.1447, simple_loss=0.1979, pruned_loss=0.04577, over 6675.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2573, pruned_loss=0.06169, over 1323636.79 frames. ], batch size: 29, lr: 1.62e-02, grad_scale: 16.0 +2023-03-20 21:07:14,993 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2916, 2.5813, 2.2172, 3.7122, 2.7611, 3.2054, 2.9829, 3.2893], + device='cuda:1'), covar=tensor([0.1524, 0.0491, 0.1873, 0.0299, 0.0029, 0.0039, 0.0080, 0.0053], + device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0222, 0.0270, 0.0238, 0.0120, 0.0121, 0.0134, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 21:07:15,996 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25919.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 21:07:26,310 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 21:07:28,456 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 21:07:28,955 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 21:07:31,490 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 21:07:36,014 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 21:07:37,887 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 +2023-03-20 21:07:40,523 INFO [train.py:901] (1/2) Epoch 10, batch 550, loss[loss=0.1894, simple_loss=0.2707, pruned_loss=0.054, over 7301.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2573, pruned_loss=0.06134, over 1348789.27 frames. ], batch size: 68, lr: 1.62e-02, grad_scale: 16.0 +2023-03-20 21:07:43,710 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1810, 1.4815, 1.3816, 1.2079, 1.3157, 0.7480, 1.0641, 0.9557], + device='cuda:1'), covar=tensor([0.0179, 0.0108, 0.0237, 0.0115, 0.0134, 0.0126, 0.0083, 0.0170], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0018, 0.0018, 0.0018, 0.0021, 0.0018, 0.0018, 0.0022], + device='cuda:1'), out_proj_covar=tensor([2.4437e-05, 2.1295e-05, 2.3237e-05, 2.1448e-05, 2.5075e-05, 2.1299e-05, + 2.1726e-05, 2.9005e-05], device='cuda:1') +2023-03-20 21:07:47,169 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 21:07:55,737 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 21:07:59,123 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 21:08:03,098 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 2.798e+02 3.426e+02 4.316e+02 1.105e+03, threshold=6.853e+02, percent-clipped=8.0 +2023-03-20 21:08:05,650 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 21:08:06,683 INFO [train.py:901] (1/2) Epoch 10, batch 600, loss[loss=0.1572, simple_loss=0.2271, pruned_loss=0.04365, over 7139.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2574, pruned_loss=0.06153, over 1369840.84 frames. ], batch size: 41, lr: 1.62e-02, grad_scale: 16.0 +2023-03-20 21:08:11,918 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3364, 0.9947, 1.1539, 1.6102, 1.1731, 1.4705, 1.1213, 1.1532], + device='cuda:1'), covar=tensor([0.0964, 0.2972, 0.1589, 0.0649, 0.4325, 0.1033, 0.1153, 0.2535], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0042, 0.0034, 0.0032, 0.0034, 0.0032, 0.0040, 0.0036], + device='cuda:1'), out_proj_covar=tensor([7.3214e-05, 9.2534e-05, 7.0756e-05, 6.7906e-05, 7.3471e-05, 7.2180e-05, + 8.5516e-05, 8.0283e-05], device='cuda:1') +2023-03-20 21:08:22,222 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 21:08:22,604 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-03-20 21:08:30,303 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4154, 4.2456, 3.8817, 3.5139, 3.9255, 2.3905, 1.6056, 4.3800], + device='cuda:1'), covar=tensor([0.0014, 0.0043, 0.0057, 0.0069, 0.0038, 0.0408, 0.0616, 0.0034], + device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0058, 0.0082, 0.0067, 0.0080, 0.0104, 0.0109, 0.0069], + device='cuda:1'), out_proj_covar=tensor([7.7435e-05, 8.9849e-05, 1.1623e-04, 9.7968e-05, 1.0912e-04, 1.4878e-04, + 1.5499e-04, 9.8657e-05], device='cuda:1') +2023-03-20 21:08:30,715 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 21:08:32,037 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 +2023-03-20 21:08:32,196 INFO [train.py:901] (1/2) Epoch 10, batch 650, loss[loss=0.1809, simple_loss=0.252, pruned_loss=0.05494, over 7332.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2583, pruned_loss=0.06151, over 1386262.40 frames. ], batch size: 61, lr: 1.62e-02, grad_scale: 16.0 +2023-03-20 21:08:38,945 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5636, 3.0829, 3.2235, 3.3336, 3.4368, 3.5750, 3.3488, 3.2472], + device='cuda:1'), covar=tensor([0.0025, 0.0079, 0.0046, 0.0044, 0.0038, 0.0032, 0.0060, 0.0059], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0036, 0.0034, 0.0031, 0.0032, 0.0032, 0.0041, 0.0041], + device='cuda:1'), out_proj_covar=tensor([7.6633e-05, 1.0907e-04, 1.0421e-04, 8.4800e-05, 8.9981e-05, 9.2383e-05, + 1.2830e-04, 1.1452e-04], device='cuda:1') +2023-03-20 21:08:41,506 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26084.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 21:08:47,362 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-20 21:08:47,952 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 21:08:49,887 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 +2023-03-20 21:08:50,560 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26102.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:08:54,997 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.614e+02 2.571e+02 3.091e+02 3.840e+02 8.546e+02, threshold=6.182e+02, percent-clipped=2.0 +2023-03-20 21:08:56,061 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 21:08:58,590 INFO [train.py:901] (1/2) Epoch 10, batch 700, loss[loss=0.2147, simple_loss=0.2747, pruned_loss=0.07735, over 6719.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2576, pruned_loss=0.06101, over 1398774.90 frames. ], batch size: 106, lr: 1.61e-02, grad_scale: 16.0 +2023-03-20 21:09:06,133 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=26132.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 21:09:16,538 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0186, 4.4725, 4.4878, 4.4155, 4.4009, 4.0195, 4.4876, 4.3890], + device='cuda:1'), covar=tensor([0.0345, 0.0318, 0.0298, 0.0381, 0.0264, 0.0321, 0.0304, 0.0423], + device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0156, 0.0117, 0.0120, 0.0101, 0.0148, 0.0132, 0.0107], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:09:19,352 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-03-20 21:09:20,060 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 21:09:20,554 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 21:09:22,251 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26163.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:09:24,092 INFO [train.py:901] (1/2) Epoch 10, batch 750, loss[loss=0.1959, simple_loss=0.2609, pruned_loss=0.06541, over 6647.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2581, pruned_loss=0.06098, over 1409498.42 frames. ], batch size: 106, lr: 1.61e-02, grad_scale: 16.0 +2023-03-20 21:09:24,751 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26168.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:09:34,168 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 21:09:38,676 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 21:09:45,707 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 21:09:46,715 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.699e+02 2.670e+02 3.310e+02 4.168e+02 9.885e+02, threshold=6.621e+02, percent-clipped=5.0 +2023-03-20 21:09:46,761 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 21:09:48,858 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26214.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 21:09:50,212 INFO [train.py:901] (1/2) Epoch 10, batch 800, loss[loss=0.2008, simple_loss=0.2579, pruned_loss=0.07185, over 7257.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2576, pruned_loss=0.0608, over 1416115.47 frames. ], batch size: 47, lr: 1.61e-02, grad_scale: 16.0 +2023-03-20 21:09:56,476 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2715, 1.7560, 2.0096, 1.5619, 1.2022, 1.1201, 1.5518, 1.2734], + device='cuda:1'), covar=tensor([0.0188, 0.0160, 0.0134, 0.0079, 0.0648, 0.0409, 0.0157, 0.0267], + device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0019, 0.0019, 0.0017, 0.0019, 0.0017, 0.0020, 0.0019], + device='cuda:1'), out_proj_covar=tensor([4.7703e-05, 4.5411e-05, 4.4031e-05, 3.7512e-05, 4.5631e-05, 4.2029e-05, + 4.5514e-05, 4.7753e-05], device='cuda:1') +2023-03-20 21:09:56,500 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26229.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:09:57,348 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 21:10:07,681 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26250.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:10:08,844 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-20 21:10:16,217 INFO [train.py:901] (1/2) Epoch 10, batch 850, loss[loss=0.1912, simple_loss=0.2646, pruned_loss=0.05891, over 7262.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2566, pruned_loss=0.0602, over 1423132.63 frames. ], batch size: 55, lr: 1.61e-02, grad_scale: 16.0 +2023-03-20 21:10:16,245 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 21:10:16,741 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 21:10:22,292 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 21:10:25,924 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 21:10:33,500 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.7979, 1.5742, 1.5248, 1.5147, 1.2256, 1.2366, 1.4288, 1.3469], + device='cuda:1'), covar=tensor([0.0380, 0.0283, 0.0225, 0.0077, 0.0515, 0.0613, 0.0260, 0.0247], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0020, 0.0020, 0.0018, 0.0020, 0.0018, 0.0020, 0.0019], + device='cuda:1'), out_proj_covar=tensor([4.9377e-05, 4.7095e-05, 4.5434e-05, 3.8040e-05, 4.7491e-05, 4.3075e-05, + 4.6891e-05, 4.8618e-05], device='cuda:1') +2023-03-20 21:10:38,437 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.623e+02 2.746e+02 3.289e+02 3.916e+02 7.980e+02, threshold=6.578e+02, percent-clipped=1.0 +2023-03-20 21:10:38,634 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6318, 3.8908, 2.4313, 4.0438, 3.3055, 4.0597, 2.3943, 2.1515], + device='cuda:1'), covar=tensor([0.0097, 0.0159, 0.1080, 0.0140, 0.0180, 0.0213, 0.1445, 0.1138], + device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0196, 0.0297, 0.0186, 0.0217, 0.0193, 0.0265, 0.0281], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 21:10:39,122 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26311.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:10:42,003 INFO [train.py:901] (1/2) Epoch 10, batch 900, loss[loss=0.2196, simple_loss=0.2788, pruned_loss=0.08018, over 7353.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2562, pruned_loss=0.05999, over 1426284.09 frames. ], batch size: 54, lr: 1.61e-02, grad_scale: 16.0 +2023-03-20 21:11:03,575 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.19 vs. limit=2.0 +2023-03-20 21:11:05,717 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 21:11:07,759 INFO [train.py:901] (1/2) Epoch 10, batch 950, loss[loss=0.2033, simple_loss=0.2688, pruned_loss=0.06891, over 7347.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.256, pruned_loss=0.0599, over 1428657.07 frames. ], batch size: 54, lr: 1.61e-02, grad_scale: 16.0 +2023-03-20 21:11:11,058 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.49 vs. limit=5.0 +2023-03-20 21:11:14,469 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0263, 1.0031, 1.1337, 1.5099, 1.1847, 1.3919, 1.3066, 1.1316], + device='cuda:1'), covar=tensor([0.2488, 0.3516, 0.0992, 0.2632, 0.2214, 0.1113, 0.1366, 0.2140], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0040, 0.0032, 0.0033, 0.0032, 0.0030, 0.0038, 0.0034], + device='cuda:1'), out_proj_covar=tensor([7.2167e-05, 8.9023e-05, 6.7553e-05, 6.9479e-05, 6.9880e-05, 7.0297e-05, + 8.3103e-05, 7.6877e-05], device='cuda:1') +2023-03-20 21:11:15,943 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0760, 4.0511, 3.5321, 3.1786, 3.6753, 2.1005, 1.6330, 3.9658], + device='cuda:1'), covar=tensor([0.0027, 0.0040, 0.0107, 0.0111, 0.0077, 0.0571, 0.0700, 0.0066], + device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0057, 0.0083, 0.0069, 0.0082, 0.0103, 0.0110, 0.0071], + device='cuda:1'), out_proj_covar=tensor([7.8325e-05, 8.8926e-05, 1.1741e-04, 9.9955e-05, 1.1263e-04, 1.4694e-04, + 1.5491e-04, 1.0076e-04], device='cuda:1') +2023-03-20 21:11:30,197 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 2.551e+02 3.073e+02 3.786e+02 9.128e+02, threshold=6.146e+02, percent-clipped=4.0 +2023-03-20 21:11:30,224 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 21:11:32,348 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26414.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:11:33,721 INFO [train.py:901] (1/2) Epoch 10, batch 1000, loss[loss=0.2199, simple_loss=0.2861, pruned_loss=0.07679, over 7252.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2565, pruned_loss=0.05999, over 1432196.18 frames. ], batch size: 55, lr: 1.61e-02, grad_scale: 16.0 +2023-03-20 21:11:47,550 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4900, 4.9476, 4.9055, 4.9142, 4.7937, 4.4552, 4.9572, 4.7955], + device='cuda:1'), covar=tensor([0.0377, 0.0339, 0.0419, 0.0402, 0.0293, 0.0292, 0.0319, 0.0470], + device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0157, 0.0119, 0.0121, 0.0104, 0.0149, 0.0132, 0.0108], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:11:50,961 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 21:11:55,014 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26458.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:11:59,400 INFO [train.py:901] (1/2) Epoch 10, batch 1050, loss[loss=0.198, simple_loss=0.2705, pruned_loss=0.06277, over 7284.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2565, pruned_loss=0.0599, over 1435839.58 frames. ], batch size: 77, lr: 1.60e-02, grad_scale: 32.0 +2023-03-20 21:12:04,244 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26475.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:12:13,040 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 21:12:16,958 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 21:12:19,016 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 +2023-03-20 21:12:22,145 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.386e+02 2.736e+02 3.241e+02 4.081e+02 9.142e+02, threshold=6.483e+02, percent-clipped=6.0 +2023-03-20 21:12:24,408 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26514.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 21:12:25,774 INFO [train.py:901] (1/2) Epoch 10, batch 1100, loss[loss=0.167, simple_loss=0.2269, pruned_loss=0.05352, over 6924.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2571, pruned_loss=0.0605, over 1438937.11 frames. ], batch size: 35, lr: 1.60e-02, grad_scale: 32.0 +2023-03-20 21:12:29,318 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26524.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:12:29,939 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8518, 3.6538, 3.3567, 3.5299, 3.0670, 2.6627, 3.8264, 2.9226], + device='cuda:1'), covar=tensor([0.0154, 0.0169, 0.0173, 0.0136, 0.0258, 0.0352, 0.0196, 0.0524], + device='cuda:1'), in_proj_covar=tensor([0.0264, 0.0262, 0.0239, 0.0254, 0.0294, 0.0288, 0.0264, 0.0289], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 21:12:31,126 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 +2023-03-20 21:12:38,608 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.24 vs. limit=5.0 +2023-03-20 21:12:46,997 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 21:12:47,520 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 21:12:49,156 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=26562.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 21:12:51,473 INFO [train.py:901] (1/2) Epoch 10, batch 1150, loss[loss=0.2041, simple_loss=0.2662, pruned_loss=0.07099, over 7349.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2561, pruned_loss=0.05994, over 1435776.71 frames. ], batch size: 73, lr: 1.60e-02, grad_scale: 32.0 +2023-03-20 21:12:53,571 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0993, 4.0880, 3.6987, 3.3235, 3.7699, 2.3439, 2.0610, 4.1245], + device='cuda:1'), covar=tensor([0.0019, 0.0052, 0.0056, 0.0071, 0.0051, 0.0418, 0.0487, 0.0044], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0058, 0.0083, 0.0070, 0.0083, 0.0105, 0.0111, 0.0073], + device='cuda:1'), out_proj_covar=tensor([7.9369e-05, 8.9778e-05, 1.1732e-04, 1.0166e-04, 1.1368e-04, 1.4962e-04, + 1.5643e-04, 1.0314e-04], device='cuda:1') +2023-03-20 21:12:55,532 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0905, 3.7280, 3.7687, 4.0272, 3.9286, 3.9942, 3.9367, 3.7521], + device='cuda:1'), covar=tensor([0.0040, 0.0071, 0.0038, 0.0033, 0.0044, 0.0041, 0.0045, 0.0073], + device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0037, 0.0034, 0.0032, 0.0034, 0.0034, 0.0042, 0.0042], + device='cuda:1'), out_proj_covar=tensor([8.0587e-05, 1.1220e-04, 1.0499e-04, 8.7419e-05, 9.4558e-05, 9.4839e-05, + 1.2904e-04, 1.1738e-04], device='cuda:1') +2023-03-20 21:12:59,464 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 21:12:59,940 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 21:13:11,816 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26606.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:13:13,757 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.942e+02 2.768e+02 3.288e+02 3.861e+02 9.785e+02, threshold=6.575e+02, percent-clipped=4.0 +2023-03-20 21:13:17,296 INFO [train.py:901] (1/2) Epoch 10, batch 1200, loss[loss=0.2202, simple_loss=0.2749, pruned_loss=0.08281, over 7297.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.257, pruned_loss=0.06086, over 1438225.60 frames. ], batch size: 86, lr: 1.60e-02, grad_scale: 32.0 +2023-03-20 21:13:32,918 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 21:13:43,156 INFO [train.py:901] (1/2) Epoch 10, batch 1250, loss[loss=0.1985, simple_loss=0.2698, pruned_loss=0.06357, over 7300.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2567, pruned_loss=0.06081, over 1438766.36 frames. ], batch size: 68, lr: 1.60e-02, grad_scale: 16.0 +2023-03-20 21:13:57,727 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 21:14:01,793 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 21:14:02,784 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 21:14:05,807 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.445e+02 2.594e+02 3.216e+02 3.895e+02 8.682e+02, threshold=6.432e+02, percent-clipped=1.0 +2023-03-20 21:14:08,779 INFO [train.py:901] (1/2) Epoch 10, batch 1300, loss[loss=0.1806, simple_loss=0.2508, pruned_loss=0.05525, over 7307.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2561, pruned_loss=0.06063, over 1437841.66 frames. ], batch size: 49, lr: 1.60e-02, grad_scale: 16.0 +2023-03-20 21:14:25,895 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 21:14:28,427 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 21:14:29,973 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26758.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:14:31,891 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 21:14:34,314 INFO [train.py:901] (1/2) Epoch 10, batch 1350, loss[loss=0.1859, simple_loss=0.2514, pruned_loss=0.06013, over 7213.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2567, pruned_loss=0.06098, over 1439384.29 frames. ], batch size: 50, lr: 1.60e-02, grad_scale: 16.0 +2023-03-20 21:14:36,521 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26770.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:14:42,416 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 21:14:54,739 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=26806.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:14:57,126 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.502e+02 2.907e+02 3.495e+02 4.130e+02 7.086e+02, threshold=6.990e+02, percent-clipped=4.0 +2023-03-20 21:14:59,191 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5653, 5.1092, 5.1574, 5.1069, 4.8062, 4.6279, 5.1811, 4.8976], + device='cuda:1'), covar=tensor([0.0398, 0.0284, 0.0302, 0.0343, 0.0373, 0.0285, 0.0274, 0.0589], + device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0162, 0.0120, 0.0122, 0.0106, 0.0149, 0.0136, 0.0110], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:15:00,105 INFO [train.py:901] (1/2) Epoch 10, batch 1400, loss[loss=0.1929, simple_loss=0.2704, pruned_loss=0.05769, over 7260.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2568, pruned_loss=0.06102, over 1439935.87 frames. ], batch size: 89, lr: 1.59e-02, grad_scale: 16.0 +2023-03-20 21:15:00,225 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5559, 2.9392, 3.2349, 3.5536, 3.5305, 3.5840, 3.1604, 3.3952], + device='cuda:1'), covar=tensor([0.0032, 0.0093, 0.0047, 0.0034, 0.0036, 0.0029, 0.0080, 0.0058], + device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0038, 0.0034, 0.0032, 0.0035, 0.0035, 0.0043, 0.0042], + device='cuda:1'), out_proj_covar=tensor([8.0122e-05, 1.1460e-04, 1.0363e-04, 8.7670e-05, 9.8476e-05, 9.7690e-05, + 1.3225e-04, 1.1883e-04], device='cuda:1') +2023-03-20 21:15:04,395 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26824.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:15:15,413 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 21:15:26,785 INFO [train.py:901] (1/2) Epoch 10, batch 1450, loss[loss=0.1965, simple_loss=0.262, pruned_loss=0.06545, over 7233.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2561, pruned_loss=0.06027, over 1441910.01 frames. ], batch size: 55, lr: 1.59e-02, grad_scale: 16.0 +2023-03-20 21:15:29,273 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=26872.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:15:33,845 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2025, 2.1388, 1.9672, 2.1595, 2.0680, 2.0963, 2.2422, 2.3322], + device='cuda:1'), covar=tensor([0.0608, 0.0768, 0.1211, 0.0529, 0.1243, 0.0723, 0.1249, 0.1302], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0033, 0.0039, 0.0036, 0.0038, 0.0036, 0.0036, 0.0034], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 21:15:39,848 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 21:15:45,365 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26904.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 21:15:46,323 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26906.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:15:49,358 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.577e+02 2.895e+02 3.554e+02 7.159e+02, threshold=5.791e+02, percent-clipped=1.0 +2023-03-20 21:15:52,289 INFO [train.py:901] (1/2) Epoch 10, batch 1500, loss[loss=0.1864, simple_loss=0.2589, pruned_loss=0.05694, over 7246.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.256, pruned_loss=0.05989, over 1442734.16 frames. ], batch size: 55, lr: 1.59e-02, grad_scale: 16.0 +2023-03-20 21:15:56,845 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 21:16:07,525 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-03-20 21:16:09,780 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26950.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:16:11,718 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=26954.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:16:11,788 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4052, 4.2466, 3.9829, 3.5976, 3.8613, 2.4978, 2.1040, 4.3559], + device='cuda:1'), covar=tensor([0.0014, 0.0033, 0.0051, 0.0045, 0.0052, 0.0332, 0.0426, 0.0032], + device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0058, 0.0080, 0.0067, 0.0082, 0.0103, 0.0109, 0.0072], + device='cuda:1'), out_proj_covar=tensor([7.8804e-05, 9.0210e-05, 1.1209e-04, 9.8218e-05, 1.1219e-04, 1.4591e-04, + 1.5358e-04, 1.0085e-04], device='cuda:1') +2023-03-20 21:16:17,368 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26965.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 21:16:18,249 INFO [train.py:901] (1/2) Epoch 10, batch 1550, loss[loss=0.1674, simple_loss=0.2438, pruned_loss=0.04552, over 7278.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2556, pruned_loss=0.05989, over 1443644.23 frames. ], batch size: 47, lr: 1.59e-02, grad_scale: 16.0 +2023-03-20 21:16:20,753 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 21:16:32,537 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 +2023-03-20 21:16:38,458 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-20 21:16:41,104 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.748e+02 2.661e+02 3.295e+02 4.115e+02 7.589e+02, threshold=6.589e+02, percent-clipped=5.0 +2023-03-20 21:16:41,267 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27011.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:16:44,152 INFO [train.py:901] (1/2) Epoch 10, batch 1600, loss[loss=0.2078, simple_loss=0.273, pruned_loss=0.07132, over 7325.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2555, pruned_loss=0.05987, over 1443089.30 frames. ], batch size: 75, lr: 1.59e-02, grad_scale: 16.0 +2023-03-20 21:16:53,172 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 21:16:53,670 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 21:16:56,658 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 21:17:05,755 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 21:17:09,740 INFO [train.py:901] (1/2) Epoch 10, batch 1650, loss[loss=0.1939, simple_loss=0.2629, pruned_loss=0.06249, over 7347.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2563, pruned_loss=0.06007, over 1443272.22 frames. ], batch size: 63, lr: 1.59e-02, grad_scale: 16.0 +2023-03-20 21:17:09,753 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 21:17:11,335 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27070.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:17:17,747 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 21:17:22,091 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5762, 3.3700, 3.1898, 3.3062, 2.6834, 2.5298, 3.5252, 2.6293], + device='cuda:1'), covar=tensor([0.0188, 0.0160, 0.0192, 0.0136, 0.0286, 0.0377, 0.0205, 0.0607], + device='cuda:1'), in_proj_covar=tensor([0.0260, 0.0265, 0.0235, 0.0255, 0.0293, 0.0287, 0.0255, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 21:17:32,624 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.597e+02 3.202e+02 3.954e+02 1.371e+03, threshold=6.405e+02, percent-clipped=3.0 +2023-03-20 21:17:35,715 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 21:17:36,197 INFO [train.py:901] (1/2) Epoch 10, batch 1700, loss[loss=0.1976, simple_loss=0.2588, pruned_loss=0.06821, over 7256.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2558, pruned_loss=0.05979, over 1442279.17 frames. ], batch size: 64, lr: 1.59e-02, grad_scale: 16.0 +2023-03-20 21:17:36,755 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=27118.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:17:39,675 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 21:17:44,320 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7324, 1.7986, 1.6901, 1.7199, 2.0821, 1.7745, 2.0002, 2.0070], + device='cuda:1'), covar=tensor([0.0748, 0.0719, 0.1363, 0.0753, 0.0554, 0.0763, 0.0789, 0.0901], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0032, 0.0038, 0.0036, 0.0038, 0.0035, 0.0034, 0.0034], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 21:17:46,238 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.2402, 5.6586, 5.7109, 5.6062, 5.3180, 5.3559, 5.7336, 5.4422], + device='cuda:1'), covar=tensor([0.0304, 0.0281, 0.0283, 0.0471, 0.0384, 0.0195, 0.0232, 0.0484], + device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0166, 0.0123, 0.0127, 0.0109, 0.0155, 0.0139, 0.0110], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:17:50,373 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 21:17:51,957 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27148.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:17:52,672 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 21:18:01,315 INFO [train.py:901] (1/2) Epoch 10, batch 1750, loss[loss=0.1897, simple_loss=0.2578, pruned_loss=0.06079, over 7257.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2559, pruned_loss=0.06, over 1441314.48 frames. ], batch size: 89, lr: 1.58e-02, grad_scale: 16.0 +2023-03-20 21:18:07,009 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3215, 3.4347, 3.2341, 3.4140, 3.3335, 3.4191, 3.7234, 3.8024], + device='cuda:1'), covar=tensor([0.0253, 0.0218, 0.0320, 0.0268, 0.0348, 0.0279, 0.0285, 0.0192], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0099, 0.0095, 0.0103, 0.0095, 0.0078, 0.0079, 0.0078], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:18:14,982 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 21:18:16,006 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 21:18:24,279 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27209.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:18:25,127 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.954e+02 2.477e+02 3.182e+02 3.732e+02 1.026e+03, threshold=6.364e+02, percent-clipped=1.0 +2023-03-20 21:18:28,117 INFO [train.py:901] (1/2) Epoch 10, batch 1800, loss[loss=0.1874, simple_loss=0.2593, pruned_loss=0.05771, over 7321.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.256, pruned_loss=0.05964, over 1442850.54 frames. ], batch size: 80, lr: 1.58e-02, grad_scale: 16.0 +2023-03-20 21:18:38,756 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 21:18:45,414 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4798, 4.8591, 4.8933, 4.8201, 4.7052, 4.4200, 4.9353, 4.7629], + device='cuda:1'), covar=tensor([0.0315, 0.0313, 0.0300, 0.0398, 0.0287, 0.0250, 0.0254, 0.0384], + device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0163, 0.0121, 0.0123, 0.0105, 0.0151, 0.0136, 0.0108], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:18:50,509 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27260.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 21:18:52,395 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 21:18:53,892 INFO [train.py:901] (1/2) Epoch 10, batch 1850, loss[loss=0.172, simple_loss=0.2296, pruned_loss=0.05724, over 6992.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2558, pruned_loss=0.05951, over 1441962.11 frames. ], batch size: 35, lr: 1.58e-02, grad_scale: 16.0 +2023-03-20 21:19:02,434 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 21:19:14,308 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27306.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:19:16,777 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.365e+02 2.681e+02 3.255e+02 3.906e+02 1.167e+03, threshold=6.510e+02, percent-clipped=5.0 +2023-03-20 21:19:19,324 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 21:19:19,783 INFO [train.py:901] (1/2) Epoch 10, batch 1900, loss[loss=0.1982, simple_loss=0.2703, pruned_loss=0.06306, over 7147.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2556, pruned_loss=0.05916, over 1444186.58 frames. ], batch size: 98, lr: 1.58e-02, grad_scale: 16.0 +2023-03-20 21:19:33,258 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.8980, 0.7892, 1.1492, 1.3517, 1.2593, 1.4084, 0.8492, 1.0513], + device='cuda:1'), covar=tensor([0.1595, 0.2230, 0.0717, 0.0467, 0.0921, 0.1179, 0.0616, 0.1903], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0041, 0.0031, 0.0035, 0.0034, 0.0033, 0.0042, 0.0037], + device='cuda:1'), out_proj_covar=tensor([7.8584e-05, 9.3853e-05, 6.8579e-05, 7.3815e-05, 7.6021e-05, 7.5430e-05, + 9.1836e-05, 8.3522e-05], device='cuda:1') +2023-03-20 21:19:40,828 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3778, 2.1731, 2.1615, 2.2584, 2.4068, 2.2806, 2.5374, 2.3829], + device='cuda:1'), covar=tensor([0.0632, 0.1190, 0.0538, 0.1316, 0.0618, 0.0374, 0.0657, 0.0613], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0033, 0.0038, 0.0036, 0.0039, 0.0035, 0.0035, 0.0035], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 21:19:42,678 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 21:19:45,114 INFO [train.py:901] (1/2) Epoch 10, batch 1950, loss[loss=0.1841, simple_loss=0.2481, pruned_loss=0.06004, over 7310.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.257, pruned_loss=0.06017, over 1444672.69 frames. ], batch size: 49, lr: 1.58e-02, grad_scale: 16.0 +2023-03-20 21:19:53,640 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 21:19:57,730 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 21:19:58,204 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 21:20:08,030 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.538e+02 2.747e+02 3.327e+02 4.450e+02 8.863e+02, threshold=6.654e+02, percent-clipped=5.0 +2023-03-20 21:20:10,962 INFO [train.py:901] (1/2) Epoch 10, batch 2000, loss[loss=0.1562, simple_loss=0.2186, pruned_loss=0.04693, over 7045.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2568, pruned_loss=0.06023, over 1443263.16 frames. ], batch size: 35, lr: 1.58e-02, grad_scale: 16.0 +2023-03-20 21:20:14,986 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 21:20:16,818 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-03-20 21:20:26,627 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 21:20:34,740 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 21:20:37,257 INFO [train.py:901] (1/2) Epoch 10, batch 2050, loss[loss=0.1886, simple_loss=0.2573, pruned_loss=0.05997, over 7360.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2563, pruned_loss=0.05998, over 1441021.97 frames. ], batch size: 51, lr: 1.58e-02, grad_scale: 16.0 +2023-03-20 21:20:42,209 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-03-20 21:20:52,733 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9786, 3.6077, 3.6597, 3.6051, 3.4483, 3.6008, 3.9179, 3.4091], + device='cuda:1'), covar=tensor([0.0114, 0.0132, 0.0139, 0.0117, 0.0263, 0.0091, 0.0114, 0.0153], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0060, 0.0061, 0.0048, 0.0086, 0.0063, 0.0062, 0.0059], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 21:20:55,684 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27504.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:20:59,079 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.657e+02 2.559e+02 3.090e+02 3.749e+02 7.460e+02, threshold=6.180e+02, percent-clipped=1.0 +2023-03-20 21:21:02,001 INFO [train.py:901] (1/2) Epoch 10, batch 2100, loss[loss=0.2111, simple_loss=0.2752, pruned_loss=0.07356, over 7338.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2563, pruned_loss=0.06023, over 1441032.76 frames. ], batch size: 63, lr: 1.57e-02, grad_scale: 16.0 +2023-03-20 21:21:07,791 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 21:21:10,874 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 21:21:14,144 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4214, 3.2860, 3.1204, 3.1145, 2.7908, 3.0720, 3.2901, 3.1336], + device='cuda:1'), covar=tensor([0.0207, 0.0260, 0.0223, 0.0297, 0.0501, 0.0209, 0.0316, 0.0212], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0060, 0.0060, 0.0048, 0.0085, 0.0062, 0.0061, 0.0057], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 21:21:25,165 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27560.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 21:21:28,513 INFO [train.py:901] (1/2) Epoch 10, batch 2150, loss[loss=0.1938, simple_loss=0.2653, pruned_loss=0.06118, over 7306.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2574, pruned_loss=0.06064, over 1441364.43 frames. ], batch size: 49, lr: 1.57e-02, grad_scale: 16.0 +2023-03-20 21:21:29,129 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8726, 2.9020, 2.9382, 3.1197, 3.2382, 3.0708, 2.4623, 2.7618], + device='cuda:1'), covar=tensor([0.1736, 0.0432, 0.1369, 0.2602, 0.0891, 0.1208, 0.1989, 0.2109], + device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0029, 0.0030, 0.0033, 0.0028, 0.0028, 0.0040, 0.0029], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 21:21:39,316 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-20 21:21:49,180 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27606.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:21:50,126 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=27608.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 21:21:51,543 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.915e+02 3.456e+02 4.146e+02 1.054e+03, threshold=6.911e+02, percent-clipped=5.0 +2023-03-20 21:21:54,473 INFO [train.py:901] (1/2) Epoch 10, batch 2200, loss[loss=0.1775, simple_loss=0.263, pruned_loss=0.04604, over 7371.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2569, pruned_loss=0.06047, over 1440250.24 frames. ], batch size: 65, lr: 1.57e-02, grad_scale: 16.0 +2023-03-20 21:21:57,124 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 21:22:08,413 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6003, 3.6180, 2.6056, 4.1739, 3.4432, 3.7773, 2.4658, 2.2081], + device='cuda:1'), covar=tensor([0.0091, 0.0650, 0.1185, 0.0134, 0.0190, 0.0179, 0.1516, 0.1267], + device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0201, 0.0295, 0.0191, 0.0221, 0.0206, 0.0266, 0.0287], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 21:22:13,769 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=27654.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:22:18,082 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-20 21:22:20,188 INFO [train.py:901] (1/2) Epoch 10, batch 2250, loss[loss=0.1967, simple_loss=0.2617, pruned_loss=0.06584, over 7274.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2559, pruned_loss=0.05986, over 1441521.92 frames. ], batch size: 70, lr: 1.57e-02, grad_scale: 16.0 +2023-03-20 21:22:30,699 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 21:22:31,193 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 21:22:36,467 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 +2023-03-20 21:22:43,305 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.716e+02 2.679e+02 3.166e+02 3.918e+02 8.017e+02, threshold=6.332e+02, percent-clipped=1.0 +2023-03-20 21:22:43,818 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 21:22:44,447 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1334, 0.9076, 1.1601, 1.5246, 1.4586, 1.3280, 0.8873, 1.2333], + device='cuda:1'), covar=tensor([0.1593, 0.2954, 0.0603, 0.0805, 0.0753, 0.0946, 0.0736, 0.1313], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0040, 0.0030, 0.0034, 0.0033, 0.0034, 0.0041, 0.0035], + device='cuda:1'), out_proj_covar=tensor([7.9296e-05, 9.2004e-05, 6.7151e-05, 7.2916e-05, 7.4398e-05, 7.6809e-05, + 8.9816e-05, 8.0813e-05], device='cuda:1') +2023-03-20 21:22:44,926 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27714.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 21:22:46,316 INFO [train.py:901] (1/2) Epoch 10, batch 2300, loss[loss=0.1886, simple_loss=0.2611, pruned_loss=0.05808, over 7305.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2562, pruned_loss=0.05985, over 1443204.62 frames. ], batch size: 80, lr: 1.57e-02, grad_scale: 16.0 +2023-03-20 21:22:47,952 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7591, 4.6643, 4.4432, 3.8734, 4.2113, 2.8982, 2.1435, 4.6315], + device='cuda:1'), covar=tensor([0.0010, 0.0053, 0.0040, 0.0046, 0.0032, 0.0290, 0.0460, 0.0029], + device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0058, 0.0081, 0.0068, 0.0083, 0.0104, 0.0110, 0.0071], + device='cuda:1'), out_proj_covar=tensor([8.0562e-05, 8.8170e-05, 1.1333e-04, 1.0012e-04, 1.1361e-04, 1.4656e-04, + 1.5520e-04, 9.9160e-05], device='cuda:1') +2023-03-20 21:22:52,790 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-03-20 21:22:56,219 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6563, 3.4318, 2.6358, 4.2803, 3.2594, 3.6515, 2.4403, 2.1903], + device='cuda:1'), covar=tensor([0.0129, 0.0309, 0.1160, 0.0178, 0.0142, 0.0508, 0.1483, 0.1235], + device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0204, 0.0298, 0.0194, 0.0221, 0.0208, 0.0268, 0.0285], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 21:23:11,759 INFO [train.py:901] (1/2) Epoch 10, batch 2350, loss[loss=0.1446, simple_loss=0.199, pruned_loss=0.04516, over 5811.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2559, pruned_loss=0.05972, over 1441248.65 frames. ], batch size: 25, lr: 1.57e-02, grad_scale: 16.0 +2023-03-20 21:23:15,933 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27775.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 21:23:16,901 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6001, 3.3272, 3.2793, 3.5228, 3.4028, 3.4717, 3.3463, 3.4877], + device='cuda:1'), covar=tensor([0.0025, 0.0071, 0.0049, 0.0039, 0.0040, 0.0039, 0.0059, 0.0056], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0038, 0.0034, 0.0033, 0.0033, 0.0035, 0.0042, 0.0042], + device='cuda:1'), out_proj_covar=tensor([7.5534e-05, 1.1539e-04, 1.0181e-04, 8.9869e-05, 8.8169e-05, 9.5921e-05, + 1.2630e-04, 1.1506e-04], device='cuda:1') +2023-03-20 21:23:27,594 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1265, 2.0210, 1.9213, 1.6289, 1.1684, 1.6979, 1.4774, 1.2984], + device='cuda:1'), covar=tensor([0.0389, 0.0103, 0.0075, 0.0078, 0.0449, 0.0187, 0.0110, 0.0265], + device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0020, 0.0021, 0.0019, 0.0020, 0.0018, 0.0021, 0.0021], + device='cuda:1'), out_proj_covar=tensor([4.9372e-05, 4.8867e-05, 4.6776e-05, 4.0536e-05, 4.8816e-05, 4.3671e-05, + 4.9377e-05, 5.2496e-05], device='cuda:1') +2023-03-20 21:23:31,435 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 21:23:32,025 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27804.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:23:32,564 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4400, 2.5300, 2.8712, 2.6808, 2.8400, 2.5539, 1.9880, 2.4828], + device='cuda:1'), covar=tensor([0.1670, 0.0424, 0.0899, 0.2709, 0.0798, 0.1182, 0.2569, 0.1670], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0029, 0.0031, 0.0032, 0.0029, 0.0028, 0.0041, 0.0030], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 21:23:35,329 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.873e+02 2.816e+02 3.299e+02 4.187e+02 9.053e+02, threshold=6.598e+02, percent-clipped=4.0 +2023-03-20 21:23:38,389 INFO [train.py:901] (1/2) Epoch 10, batch 2400, loss[loss=0.1748, simple_loss=0.22, pruned_loss=0.06481, over 6201.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2557, pruned_loss=0.05942, over 1441124.42 frames. ], batch size: 26, lr: 1.57e-02, grad_scale: 16.0 +2023-03-20 21:23:38,393 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 21:23:48,935 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 21:23:51,455 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 21:23:55,995 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=27852.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:24:03,558 INFO [train.py:901] (1/2) Epoch 10, batch 2450, loss[loss=0.1882, simple_loss=0.2601, pruned_loss=0.05819, over 7222.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2545, pruned_loss=0.05915, over 1440911.19 frames. ], batch size: 93, lr: 1.56e-02, grad_scale: 16.0 +2023-03-20 21:24:12,590 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-03-20 21:24:19,846 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 21:24:26,790 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.336e+02 2.708e+02 3.245e+02 4.079e+02 9.497e+02, threshold=6.490e+02, percent-clipped=3.0 +2023-03-20 21:24:29,834 INFO [train.py:901] (1/2) Epoch 10, batch 2500, loss[loss=0.1632, simple_loss=0.2286, pruned_loss=0.04888, over 6970.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.255, pruned_loss=0.05924, over 1440995.90 frames. ], batch size: 35, lr: 1.56e-02, grad_scale: 16.0 +2023-03-20 21:24:31,971 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9787, 4.1441, 3.7738, 4.1512, 3.9399, 4.0684, 4.3770, 4.3716], + device='cuda:1'), covar=tensor([0.0198, 0.0133, 0.0205, 0.0127, 0.0271, 0.0182, 0.0246, 0.0198], + device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0096, 0.0090, 0.0097, 0.0093, 0.0078, 0.0075, 0.0077], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:24:43,368 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 21:24:55,328 INFO [train.py:901] (1/2) Epoch 10, batch 2550, loss[loss=0.1586, simple_loss=0.2311, pruned_loss=0.0431, over 7280.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2544, pruned_loss=0.05887, over 1442267.26 frames. ], batch size: 47, lr: 1.56e-02, grad_scale: 16.0 +2023-03-20 21:25:21,732 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 +2023-03-20 21:25:21,855 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.650e+02 2.672e+02 3.067e+02 3.861e+02 6.346e+02, threshold=6.134e+02, percent-clipped=0.0 +2023-03-20 21:25:24,834 INFO [train.py:901] (1/2) Epoch 10, batch 2600, loss[loss=0.2172, simple_loss=0.2827, pruned_loss=0.07584, over 7216.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2549, pruned_loss=0.05911, over 1442495.33 frames. ], batch size: 93, lr: 1.56e-02, grad_scale: 16.0 +2023-03-20 21:25:49,407 INFO [train.py:901] (1/2) Epoch 10, batch 2650, loss[loss=0.1795, simple_loss=0.2519, pruned_loss=0.05357, over 7270.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2559, pruned_loss=0.05954, over 1442685.89 frames. ], batch size: 77, lr: 1.56e-02, grad_scale: 16.0 +2023-03-20 21:25:50,571 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0061, 3.6047, 3.3873, 3.7237, 2.7755, 2.6099, 4.0938, 3.0988], + device='cuda:1'), covar=tensor([0.0138, 0.0146, 0.0276, 0.0196, 0.0447, 0.0580, 0.0186, 0.0797], + device='cuda:1'), in_proj_covar=tensor([0.0265, 0.0267, 0.0234, 0.0266, 0.0297, 0.0289, 0.0264, 0.0291], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 21:25:50,970 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28070.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 21:26:05,340 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7618, 3.8730, 3.8357, 3.9630, 3.5743, 3.9330, 4.2112, 4.2521], + device='cuda:1'), covar=tensor([0.0202, 0.0166, 0.0166, 0.0127, 0.0322, 0.0198, 0.0198, 0.0153], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0097, 0.0089, 0.0097, 0.0095, 0.0078, 0.0076, 0.0076], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:26:05,378 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7834, 2.4528, 2.9504, 2.9000, 3.0884, 2.5084, 2.2010, 2.8010], + device='cuda:1'), covar=tensor([0.1117, 0.0360, 0.1323, 0.1237, 0.0961, 0.2095, 0.2835, 0.1459], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0029, 0.0029, 0.0030, 0.0027, 0.0027, 0.0039, 0.0028], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 21:26:11,632 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.564e+02 2.458e+02 3.041e+02 3.552e+02 7.136e+02, threshold=6.082e+02, percent-clipped=4.0 +2023-03-20 21:26:13,414 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.02 vs. limit=5.0 +2023-03-20 21:26:14,555 INFO [train.py:901] (1/2) Epoch 10, batch 2700, loss[loss=0.1868, simple_loss=0.2573, pruned_loss=0.05811, over 7286.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2553, pruned_loss=0.05935, over 1443745.52 frames. ], batch size: 57, lr: 1.56e-02, grad_scale: 16.0 +2023-03-20 21:26:16,159 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6159, 2.1674, 2.7618, 2.7559, 2.7184, 2.2336, 1.8864, 2.6720], + device='cuda:1'), covar=tensor([0.1467, 0.0484, 0.1095, 0.1034, 0.1236, 0.2486, 0.2833, 0.1246], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0029, 0.0029, 0.0030, 0.0027, 0.0027, 0.0039, 0.0027], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 21:26:25,563 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-03-20 21:26:39,397 INFO [train.py:901] (1/2) Epoch 10, batch 2750, loss[loss=0.1881, simple_loss=0.2587, pruned_loss=0.05875, over 7335.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2544, pruned_loss=0.0589, over 1444948.10 frames. ], batch size: 61, lr: 1.56e-02, grad_scale: 16.0 +2023-03-20 21:26:48,158 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-03-20 21:27:01,429 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.622e+02 3.116e+02 3.809e+02 8.984e+02, threshold=6.231e+02, percent-clipped=2.0 +2023-03-20 21:27:04,488 INFO [train.py:901] (1/2) Epoch 10, batch 2800, loss[loss=0.1917, simple_loss=0.273, pruned_loss=0.05523, over 7320.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2539, pruned_loss=0.05863, over 1443144.77 frames. ], batch size: 83, lr: 1.56e-02, grad_scale: 16.0 +2023-03-20 21:27:10,532 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28229.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:27:30,257 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 21:27:30,313 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 21:27:30,339 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 21:27:30,356 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 21:27:30,532 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 21:27:30,549 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 21:27:30,560 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 21:27:30,684 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 21:27:30,763 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 21:27:30,778 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 21:27:30,796 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 21:27:31,307 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 21:27:31,314 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 21:27:31,321 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 21:27:31,464 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 21:27:31,467 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 21:27:31,528 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 21:27:39,081 INFO [train.py:901] (1/2) Epoch 11, batch 0, loss[loss=0.201, simple_loss=0.2801, pruned_loss=0.06095, over 7139.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2801, pruned_loss=0.06095, over 7139.00 frames. ], batch size: 98, lr: 1.50e-02, grad_scale: 16.0 +2023-03-20 21:27:39,081 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 21:27:44,691 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3100, 2.1845, 2.0667, 3.0019, 1.5539, 2.9215, 1.3168, 2.6518], + device='cuda:1'), covar=tensor([0.0059, 0.0893, 0.1625, 0.0067, 0.4527, 0.0087, 0.1183, 0.0134], + device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0267, 0.0311, 0.0141, 0.0296, 0.0149, 0.0274, 0.0192], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 21:27:52,677 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0814, 1.3324, 1.0347, 1.1710, 0.9590, 0.9239, 1.1313, 0.9505], + device='cuda:1'), covar=tensor([0.0208, 0.0113, 0.0246, 0.0217, 0.0326, 0.0114, 0.0198, 0.0191], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0018, 0.0018, 0.0019, 0.0021, 0.0019, 0.0018, 0.0023], + device='cuda:1'), out_proj_covar=tensor([2.3543e-05, 2.0855e-05, 2.2578e-05, 2.1583e-05, 2.6241e-05, 2.1273e-05, + 2.1386e-05, 2.9579e-05], device='cuda:1') +2023-03-20 21:28:04,270 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2112, 1.4146, 1.2857, 1.1529, 1.0686, 0.9183, 1.3089, 0.9847], + device='cuda:1'), covar=tensor([0.0168, 0.0086, 0.0155, 0.0395, 0.0703, 0.0167, 0.0188, 0.0232], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0018, 0.0018, 0.0019, 0.0021, 0.0019, 0.0018, 0.0023], + device='cuda:1'), out_proj_covar=tensor([2.3543e-05, 2.0855e-05, 2.2578e-05, 2.1583e-05, 2.6241e-05, 2.1273e-05, + 2.1386e-05, 2.9579e-05], device='cuda:1') +2023-03-20 21:28:04,929 INFO [train.py:935] (1/2) Epoch 11, validation: loss=0.1751, simple_loss=0.2617, pruned_loss=0.04426, over 1622729.00 frames. +2023-03-20 21:28:04,930 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-20 21:28:11,459 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 21:28:11,566 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7154, 3.8582, 3.5980, 3.7346, 3.4812, 3.7348, 4.0932, 4.1622], + device='cuda:1'), covar=tensor([0.0231, 0.0144, 0.0242, 0.0242, 0.0406, 0.0256, 0.0245, 0.0213], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0097, 0.0091, 0.0100, 0.0094, 0.0079, 0.0076, 0.0078], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:28:12,111 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28255.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 21:28:21,927 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 21:28:22,786 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-03-20 21:28:23,493 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4738, 4.9344, 5.0231, 4.8991, 4.7532, 4.5140, 4.9880, 4.8438], + device='cuda:1'), covar=tensor([0.0350, 0.0274, 0.0225, 0.0322, 0.0248, 0.0232, 0.0264, 0.0352], + device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0163, 0.0121, 0.0119, 0.0103, 0.0150, 0.0135, 0.0106], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:28:28,044 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28287.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:28:28,475 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 21:28:29,592 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28290.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:28:29,922 INFO [train.py:901] (1/2) Epoch 11, batch 50, loss[loss=0.1595, simple_loss=0.241, pruned_loss=0.03897, over 7264.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.259, pruned_loss=0.06025, over 325477.60 frames. ], batch size: 89, lr: 1.50e-02, grad_scale: 16.0 +2023-03-20 21:28:30,419 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 21:28:32,929 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 21:28:38,053 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-20 21:28:39,541 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-20 21:28:40,668 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.663e+02 3.200e+02 3.630e+02 7.149e+02, threshold=6.399e+02, percent-clipped=1.0 +2023-03-20 21:28:43,281 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28316.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 21:28:51,399 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 21:28:51,870 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 21:28:56,458 INFO [train.py:901] (1/2) Epoch 11, batch 100, loss[loss=0.1604, simple_loss=0.2384, pruned_loss=0.04118, over 7314.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2554, pruned_loss=0.05906, over 572369.14 frames. ], batch size: 59, lr: 1.49e-02, grad_scale: 16.0 +2023-03-20 21:29:00,148 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28348.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:29:11,327 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28370.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 21:29:21,752 INFO [train.py:901] (1/2) Epoch 11, batch 150, loss[loss=0.2081, simple_loss=0.2731, pruned_loss=0.07155, over 7242.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2541, pruned_loss=0.05823, over 764018.84 frames. ], batch size: 93, lr: 1.49e-02, grad_scale: 16.0 +2023-03-20 21:29:32,723 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.668e+02 3.214e+02 3.934e+02 8.081e+02, threshold=6.428e+02, percent-clipped=3.0 +2023-03-20 21:29:33,047 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 +2023-03-20 21:29:36,377 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=28418.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 21:29:48,523 INFO [train.py:901] (1/2) Epoch 11, batch 200, loss[loss=0.184, simple_loss=0.2538, pruned_loss=0.05706, over 7256.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2542, pruned_loss=0.05745, over 917070.79 frames. ], batch size: 55, lr: 1.49e-02, grad_scale: 16.0 +2023-03-20 21:29:54,547 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 21:29:59,675 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 21:30:05,852 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 21:30:12,014 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28486.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:30:14,381 INFO [train.py:901] (1/2) Epoch 11, batch 250, loss[loss=0.1739, simple_loss=0.2548, pruned_loss=0.04649, over 7248.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2536, pruned_loss=0.05764, over 1034384.38 frames. ], batch size: 55, lr: 1.49e-02, grad_scale: 16.0 +2023-03-20 21:30:15,025 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28492.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:30:18,983 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 21:30:25,035 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.444e+02 2.592e+02 3.063e+02 3.811e+02 9.501e+02, threshold=6.125e+02, percent-clipped=2.0 +2023-03-20 21:30:30,141 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0289, 3.6830, 3.6815, 3.6126, 3.4962, 3.7934, 3.9261, 3.6169], + device='cuda:1'), covar=tensor([0.0076, 0.0132, 0.0131, 0.0147, 0.0243, 0.0063, 0.0126, 0.0104], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0060, 0.0061, 0.0050, 0.0089, 0.0063, 0.0063, 0.0058], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 21:30:34,831 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-03-20 21:30:40,039 INFO [train.py:901] (1/2) Epoch 11, batch 300, loss[loss=0.2026, simple_loss=0.2644, pruned_loss=0.07041, over 7279.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2534, pruned_loss=0.05763, over 1123792.93 frames. ], batch size: 66, lr: 1.49e-02, grad_scale: 16.0 +2023-03-20 21:30:40,062 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 21:30:43,224 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28547.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:30:46,227 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28553.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:30:48,086 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 21:30:48,920 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 +2023-03-20 21:31:02,798 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28585.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:31:05,758 INFO [train.py:901] (1/2) Epoch 11, batch 350, loss[loss=0.1654, simple_loss=0.236, pruned_loss=0.04743, over 6982.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2542, pruned_loss=0.05784, over 1193759.28 frames. ], batch size: 35, lr: 1.49e-02, grad_scale: 16.0 +2023-03-20 21:31:16,858 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.554e+02 2.672e+02 3.200e+02 3.823e+02 8.890e+02, threshold=6.400e+02, percent-clipped=4.0 +2023-03-20 21:31:16,960 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28611.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 21:31:22,953 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 21:31:23,029 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.2533, 4.8760, 4.8574, 5.2024, 5.2412, 5.2697, 4.6973, 4.8286], + device='cuda:1'), covar=tensor([0.0590, 0.1977, 0.1609, 0.1074, 0.0622, 0.1017, 0.0690, 0.0769], + device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0284, 0.0230, 0.0223, 0.0173, 0.0279, 0.0165, 0.0198], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:31:32,150 INFO [train.py:901] (1/2) Epoch 11, batch 400, loss[loss=0.1504, simple_loss=0.2245, pruned_loss=0.03818, over 7281.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2534, pruned_loss=0.05753, over 1249200.58 frames. ], batch size: 42, lr: 1.49e-02, grad_scale: 16.0 +2023-03-20 21:31:33,283 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28643.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:31:40,998 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6225, 2.5076, 2.8903, 2.6529, 2.6353, 2.6105, 2.1196, 2.4594], + device='cuda:1'), covar=tensor([0.1979, 0.0662, 0.1570, 0.2186, 0.1202, 0.1583, 0.3172, 0.2540], + device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0031, 0.0031, 0.0032, 0.0028, 0.0029, 0.0041, 0.0029], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 21:31:47,941 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8705, 3.9129, 3.2867, 3.1252, 3.3989, 2.2444, 1.8820, 3.9359], + device='cuda:1'), covar=tensor([0.0023, 0.0031, 0.0082, 0.0053, 0.0070, 0.0365, 0.0442, 0.0028], + device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0058, 0.0079, 0.0067, 0.0082, 0.0103, 0.0109, 0.0069], + device='cuda:1'), out_proj_covar=tensor([8.0432e-05, 8.7806e-05, 1.0990e-04, 9.8600e-05, 1.1271e-04, 1.4542e-04, + 1.5202e-04, 9.5371e-05], device='cuda:1') +2023-03-20 21:31:58,569 INFO [train.py:901] (1/2) Epoch 11, batch 450, loss[loss=0.171, simple_loss=0.2424, pruned_loss=0.04974, over 7306.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2543, pruned_loss=0.05777, over 1293012.46 frames. ], batch size: 59, lr: 1.49e-02, grad_scale: 32.0 +2023-03-20 21:32:01,973 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 +2023-03-20 21:32:05,672 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 21:32:06,154 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 21:32:08,671 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.433e+02 3.005e+02 3.670e+02 6.203e+02, threshold=6.010e+02, percent-clipped=0.0 +2023-03-20 21:32:23,920 INFO [train.py:901] (1/2) Epoch 11, batch 500, loss[loss=0.1729, simple_loss=0.234, pruned_loss=0.05588, over 6997.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.254, pruned_loss=0.05746, over 1325940.46 frames. ], batch size: 35, lr: 1.48e-02, grad_scale: 16.0 +2023-03-20 21:32:25,214 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 +2023-03-20 21:32:34,550 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2546, 4.3288, 4.1490, 4.2910, 3.9581, 4.4099, 4.6325, 4.6801], + device='cuda:1'), covar=tensor([0.0168, 0.0136, 0.0184, 0.0158, 0.0344, 0.0130, 0.0180, 0.0159], + device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0097, 0.0089, 0.0101, 0.0095, 0.0076, 0.0076, 0.0079], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:32:39,599 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 21:32:41,077 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 21:32:42,081 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 21:32:44,593 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 21:32:48,798 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2909, 1.7887, 2.0185, 1.5469, 1.0190, 1.5145, 1.5994, 1.3049], + device='cuda:1'), covar=tensor([0.0294, 0.0232, 0.0123, 0.0145, 0.0364, 0.0286, 0.0198, 0.0354], + device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0020, 0.0020, 0.0019, 0.0019, 0.0016, 0.0022, 0.0020], + device='cuda:1'), out_proj_covar=tensor([4.7663e-05, 4.8259e-05, 4.5613e-05, 3.9883e-05, 4.6903e-05, 4.0980e-05, + 5.0429e-05, 4.9294e-05], device='cuda:1') +2023-03-20 21:32:49,693 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 21:32:50,217 INFO [train.py:901] (1/2) Epoch 11, batch 550, loss[loss=0.1793, simple_loss=0.2517, pruned_loss=0.05342, over 7337.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2522, pruned_loss=0.0569, over 1350400.15 frames. ], batch size: 61, lr: 1.48e-02, grad_scale: 16.0 +2023-03-20 21:33:00,822 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 21:33:01,104 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 +2023-03-20 21:33:01,277 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.650e+02 3.021e+02 3.738e+02 7.326e+02, threshold=6.042e+02, percent-clipped=1.0 +2023-03-20 21:33:06,446 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3603, 2.5125, 2.1123, 2.1036, 2.4523, 2.2565, 2.6553, 2.4749], + device='cuda:1'), covar=tensor([0.2268, 0.1352, 0.1405, 0.1661, 0.2638, 0.0957, 0.0526, 0.0978], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0033, 0.0040, 0.0036, 0.0040, 0.0037, 0.0036, 0.0036], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 21:33:08,859 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 21:33:12,893 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 21:33:16,510 INFO [train.py:901] (1/2) Epoch 11, batch 600, loss[loss=0.1719, simple_loss=0.241, pruned_loss=0.05142, over 7162.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2517, pruned_loss=0.05637, over 1371508.96 frames. ], batch size: 41, lr: 1.48e-02, grad_scale: 16.0 +2023-03-20 21:33:17,095 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28842.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:33:19,525 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 21:33:20,081 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28848.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:33:27,395 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28861.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:33:32,958 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3159, 1.8291, 2.1140, 1.3948, 1.2184, 1.4841, 1.5774, 1.3887], + device='cuda:1'), covar=tensor([0.0300, 0.0177, 0.0073, 0.0123, 0.0318, 0.0336, 0.0156, 0.0173], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0020, 0.0020, 0.0019, 0.0018, 0.0016, 0.0021, 0.0019], + device='cuda:1'), out_proj_covar=tensor([4.6806e-05, 4.8031e-05, 4.4927e-05, 3.9702e-05, 4.5863e-05, 4.0159e-05, + 4.9420e-05, 4.8526e-05], device='cuda:1') +2023-03-20 21:33:36,265 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 21:33:39,295 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28885.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:33:42,191 INFO [train.py:901] (1/2) Epoch 11, batch 650, loss[loss=0.1428, simple_loss=0.2093, pruned_loss=0.03814, over 6990.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2519, pruned_loss=0.05637, over 1388700.52 frames. ], batch size: 35, lr: 1.48e-02, grad_scale: 16.0 +2023-03-20 21:33:45,180 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 21:33:52,341 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28911.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 21:33:52,676 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.509e+02 2.625e+02 3.059e+02 3.685e+02 8.000e+02, threshold=6.119e+02, percent-clipped=1.0 +2023-03-20 21:33:58,576 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28922.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:34:01,956 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 21:34:03,924 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=28933.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:34:07,921 INFO [train.py:901] (1/2) Epoch 11, batch 700, loss[loss=0.1412, simple_loss=0.1967, pruned_loss=0.04285, over 5862.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2528, pruned_loss=0.05674, over 1401062.91 frames. ], batch size: 25, lr: 1.48e-02, grad_scale: 16.0 +2023-03-20 21:34:09,721 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28943.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:34:11,140 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 21:34:17,774 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=28959.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 21:34:17,894 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2827, 3.1337, 2.8057, 3.2160, 2.5235, 2.2225, 3.4326, 2.5581], + device='cuda:1'), covar=tensor([0.0243, 0.0154, 0.0252, 0.0184, 0.0305, 0.0430, 0.0241, 0.0636], + device='cuda:1'), in_proj_covar=tensor([0.0264, 0.0270, 0.0236, 0.0268, 0.0293, 0.0290, 0.0267, 0.0291], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 21:34:33,893 INFO [train.py:901] (1/2) Epoch 11, batch 750, loss[loss=0.149, simple_loss=0.2164, pruned_loss=0.04084, over 7041.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2528, pruned_loss=0.05648, over 1407428.57 frames. ], batch size: 35, lr: 1.48e-02, grad_scale: 16.0 +2023-03-20 21:34:33,963 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=28991.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:34:34,408 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 21:34:34,965 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 21:34:40,583 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29002.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:34:45,597 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.567e+02 3.046e+02 3.773e+02 7.395e+02, threshold=6.092e+02, percent-clipped=4.0 +2023-03-20 21:34:50,179 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 21:34:55,272 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 21:35:00,677 INFO [train.py:901] (1/2) Epoch 11, batch 800, loss[loss=0.1853, simple_loss=0.2431, pruned_loss=0.06379, over 7339.00 frames. ], tot_loss[loss=0.182, simple_loss=0.252, pruned_loss=0.05604, over 1413991.05 frames. ], batch size: 44, lr: 1.48e-02, grad_scale: 16.0 +2023-03-20 21:35:00,700 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 21:35:01,506 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.02 vs. limit=2.0 +2023-03-20 21:35:02,194 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 21:35:02,366 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4302, 3.2187, 2.9816, 3.2322, 2.4504, 2.3191, 3.4392, 2.5420], + device='cuda:1'), covar=tensor([0.0156, 0.0154, 0.0223, 0.0162, 0.0329, 0.0462, 0.0248, 0.0628], + device='cuda:1'), in_proj_covar=tensor([0.0264, 0.0275, 0.0234, 0.0269, 0.0294, 0.0291, 0.0269, 0.0291], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 21:35:11,991 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1026, 1.3748, 1.0854, 1.3783, 1.3148, 1.0273, 0.8360, 0.8764], + device='cuda:1'), covar=tensor([0.0136, 0.0137, 0.0236, 0.0121, 0.0123, 0.0085, 0.0177, 0.0144], + device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0019, 0.0019, 0.0018, 0.0022, 0.0019, 0.0020, 0.0024], + device='cuda:1'), out_proj_covar=tensor([2.4190e-05, 2.2469e-05, 2.3566e-05, 2.0726e-05, 2.6593e-05, 2.1652e-05, + 2.3746e-05, 3.0124e-05], device='cuda:1') +2023-03-20 21:35:11,996 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29063.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:35:12,380 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 21:35:13,656 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 +2023-03-20 21:35:25,734 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2422, 1.8027, 1.9796, 1.3748, 1.2479, 1.3107, 1.3754, 1.4652], + device='cuda:1'), covar=tensor([0.0371, 0.0107, 0.0111, 0.0070, 0.0323, 0.0261, 0.0192, 0.0134], + device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0021, 0.0021, 0.0019, 0.0019, 0.0017, 0.0022, 0.0020], + device='cuda:1'), out_proj_covar=tensor([4.8916e-05, 5.0947e-05, 4.7476e-05, 4.0289e-05, 4.8225e-05, 4.2210e-05, + 5.1673e-05, 4.9526e-05], device='cuda:1') +2023-03-20 21:35:26,644 INFO [train.py:901] (1/2) Epoch 11, batch 850, loss[loss=0.1935, simple_loss=0.2584, pruned_loss=0.06427, over 7241.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2521, pruned_loss=0.05584, over 1421994.91 frames. ], batch size: 55, lr: 1.48e-02, grad_scale: 16.0 +2023-03-20 21:35:30,430 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2375, 1.6257, 2.0349, 1.4584, 1.3224, 1.3291, 1.4219, 1.4350], + device='cuda:1'), covar=tensor([0.0358, 0.0356, 0.0168, 0.0085, 0.0496, 0.0299, 0.0334, 0.0196], + device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0021, 0.0021, 0.0019, 0.0019, 0.0017, 0.0023, 0.0020], + device='cuda:1'), out_proj_covar=tensor([4.9142e-05, 5.1126e-05, 4.7843e-05, 4.0268e-05, 4.8112e-05, 4.2424e-05, + 5.2056e-05, 4.9584e-05], device='cuda:1') +2023-03-20 21:35:31,866 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 21:35:32,402 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 21:35:37,485 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.474e+02 2.954e+02 3.658e+02 5.777e+02, threshold=5.907e+02, percent-clipped=0.0 +2023-03-20 21:35:37,539 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 21:35:41,651 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 21:35:52,507 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 +2023-03-20 21:35:52,627 INFO [train.py:901] (1/2) Epoch 11, batch 900, loss[loss=0.2097, simple_loss=0.2828, pruned_loss=0.06826, over 7279.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2516, pruned_loss=0.05566, over 1426474.04 frames. ], batch size: 86, lr: 1.47e-02, grad_scale: 16.0 +2023-03-20 21:35:53,222 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29142.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:35:56,302 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29148.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:36:06,089 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.39 vs. limit=5.0 +2023-03-20 21:36:17,981 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=29190.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:36:18,430 INFO [train.py:901] (1/2) Epoch 11, batch 950, loss[loss=0.1792, simple_loss=0.2527, pruned_loss=0.0528, over 7317.00 frames. ], tot_loss[loss=0.182, simple_loss=0.252, pruned_loss=0.05597, over 1430930.13 frames. ], batch size: 83, lr: 1.47e-02, grad_scale: 16.0 +2023-03-20 21:36:19,427 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 21:36:21,019 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=29196.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:36:30,011 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.761e+02 2.655e+02 3.153e+02 4.188e+02 9.756e+02, threshold=6.306e+02, percent-clipped=6.0 +2023-03-20 21:36:32,638 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29217.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:36:42,193 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29236.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:36:44,571 INFO [train.py:901] (1/2) Epoch 11, batch 1000, loss[loss=0.1591, simple_loss=0.2321, pruned_loss=0.04311, over 7356.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2524, pruned_loss=0.05613, over 1432691.31 frames. ], batch size: 44, lr: 1.47e-02, grad_scale: 16.0 +2023-03-20 21:36:44,594 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 21:37:05,453 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 21:37:07,125 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1267, 2.6315, 2.3798, 3.5859, 1.5570, 3.5297, 1.4582, 3.0463], + device='cuda:1'), covar=tensor([0.0035, 0.0840, 0.1576, 0.0064, 0.4273, 0.0061, 0.1309, 0.0186], + device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0264, 0.0303, 0.0141, 0.0295, 0.0145, 0.0267, 0.0190], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 21:37:11,088 INFO [train.py:901] (1/2) Epoch 11, batch 1050, loss[loss=0.1696, simple_loss=0.2342, pruned_loss=0.05251, over 7180.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2526, pruned_loss=0.05646, over 1435039.78 frames. ], batch size: 39, lr: 1.47e-02, grad_scale: 16.0 +2023-03-20 21:37:14,293 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29297.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:37:21,636 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.500e+02 3.009e+02 3.529e+02 7.221e+02, threshold=6.018e+02, percent-clipped=4.0 +2023-03-20 21:37:27,693 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 21:37:31,230 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 21:37:36,259 INFO [train.py:901] (1/2) Epoch 11, batch 1100, loss[loss=0.1773, simple_loss=0.2502, pruned_loss=0.05223, over 7269.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2527, pruned_loss=0.05633, over 1436298.90 frames. ], batch size: 57, lr: 1.47e-02, grad_scale: 16.0 +2023-03-20 21:37:45,442 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29358.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:38:00,564 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 21:38:01,083 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 21:38:02,613 INFO [train.py:901] (1/2) Epoch 11, batch 1150, loss[loss=0.1854, simple_loss=0.2568, pruned_loss=0.05705, over 7345.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2527, pruned_loss=0.05642, over 1439476.19 frames. ], batch size: 75, lr: 1.47e-02, grad_scale: 16.0 +2023-03-20 21:38:13,506 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.646e+02 2.488e+02 2.965e+02 3.681e+02 7.032e+02, threshold=5.930e+02, percent-clipped=2.0 +2023-03-20 21:38:13,554 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 21:38:14,548 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 21:38:25,811 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8064, 1.9899, 1.6229, 3.0457, 2.7356, 2.9425, 2.5736, 2.2598], + device='cuda:1'), covar=tensor([0.1490, 0.0783, 0.2415, 0.0605, 0.0076, 0.0057, 0.0073, 0.0070], + device='cuda:1'), in_proj_covar=tensor([0.0291, 0.0222, 0.0274, 0.0243, 0.0123, 0.0115, 0.0138, 0.0151], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], + device='cuda:1') +2023-03-20 21:38:28,832 INFO [train.py:901] (1/2) Epoch 11, batch 1200, loss[loss=0.1748, simple_loss=0.2503, pruned_loss=0.0497, over 7344.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2541, pruned_loss=0.05708, over 1442642.54 frames. ], batch size: 54, lr: 1.47e-02, grad_scale: 16.0 +2023-03-20 21:38:35,244 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 +2023-03-20 21:38:47,615 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 21:38:54,676 INFO [train.py:901] (1/2) Epoch 11, batch 1250, loss[loss=0.1843, simple_loss=0.253, pruned_loss=0.05776, over 7309.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2533, pruned_loss=0.05676, over 1441274.59 frames. ], batch size: 86, lr: 1.47e-02, grad_scale: 16.0 +2023-03-20 21:39:04,976 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0474, 1.7469, 1.9131, 1.2578, 1.6373, 1.3535, 1.1944, 1.3661], + device='cuda:1'), covar=tensor([0.0382, 0.0161, 0.0156, 0.0086, 0.0194, 0.0261, 0.0246, 0.0263], + device='cuda:1'), in_proj_covar=tensor([0.0019, 0.0019, 0.0020, 0.0018, 0.0018, 0.0015, 0.0020, 0.0018], + device='cuda:1'), out_proj_covar=tensor([4.7405e-05, 4.6084e-05, 4.5249e-05, 3.8643e-05, 4.4542e-05, 3.9432e-05, + 4.7655e-05, 4.6932e-05], device='cuda:1') +2023-03-20 21:39:05,276 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.616e+02 2.504e+02 3.020e+02 3.915e+02 7.336e+02, threshold=6.040e+02, percent-clipped=4.0 +2023-03-20 21:39:07,930 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29517.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:39:12,628 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-03-20 21:39:12,858 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 21:39:16,992 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 21:39:17,857 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-20 21:39:18,001 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 21:39:20,481 INFO [train.py:901] (1/2) Epoch 11, batch 1300, loss[loss=0.1783, simple_loss=0.2406, pruned_loss=0.05797, over 7204.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.252, pruned_loss=0.05653, over 1438707.24 frames. ], batch size: 50, lr: 1.47e-02, grad_scale: 8.0 +2023-03-20 21:39:26,248 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29551.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:39:28,471 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 +2023-03-20 21:39:33,262 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=29565.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:39:41,753 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 21:39:43,809 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 21:39:46,289 INFO [train.py:901] (1/2) Epoch 11, batch 1350, loss[loss=0.1986, simple_loss=0.2635, pruned_loss=0.06691, over 7350.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2522, pruned_loss=0.05653, over 1441020.03 frames. ], batch size: 73, lr: 1.46e-02, grad_scale: 8.0 +2023-03-20 21:39:46,863 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29592.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:39:47,318 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 21:39:57,329 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29612.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:39:57,659 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 2.822e+02 3.499e+02 4.170e+02 8.091e+02, threshold=6.998e+02, percent-clipped=6.0 +2023-03-20 21:39:58,187 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 21:40:12,596 INFO [train.py:901] (1/2) Epoch 11, batch 1400, loss[loss=0.1809, simple_loss=0.2512, pruned_loss=0.05528, over 7278.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2517, pruned_loss=0.05632, over 1441163.21 frames. ], batch size: 66, lr: 1.46e-02, grad_scale: 8.0 +2023-03-20 21:40:21,789 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29658.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:40:23,788 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29662.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:40:33,106 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 21:40:38,172 INFO [train.py:901] (1/2) Epoch 11, batch 1450, loss[loss=0.1932, simple_loss=0.2682, pruned_loss=0.05907, over 7264.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2513, pruned_loss=0.05601, over 1442182.76 frames. ], batch size: 70, lr: 1.46e-02, grad_scale: 8.0 +2023-03-20 21:40:46,585 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=29706.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:40:49,930 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.574e+02 2.541e+02 3.078e+02 3.922e+02 1.429e+03, threshold=6.155e+02, percent-clipped=1.0 +2023-03-20 21:40:52,495 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7568, 4.0122, 4.3874, 4.2164, 4.0685, 4.3301, 4.5416, 4.1854], + device='cuda:1'), covar=tensor([0.0088, 0.0129, 0.0111, 0.0107, 0.0248, 0.0068, 0.0129, 0.0088], + device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0063, 0.0065, 0.0052, 0.0096, 0.0068, 0.0065, 0.0063], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:40:55,051 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29723.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 21:40:57,407 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 21:41:04,512 INFO [train.py:901] (1/2) Epoch 11, batch 1500, loss[loss=0.1791, simple_loss=0.2434, pruned_loss=0.0574, over 7288.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2515, pruned_loss=0.05652, over 1440534.29 frames. ], batch size: 42, lr: 1.46e-02, grad_scale: 8.0 +2023-03-20 21:41:13,019 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 21:41:30,361 INFO [train.py:901] (1/2) Epoch 11, batch 1550, loss[loss=0.1903, simple_loss=0.2591, pruned_loss=0.06073, over 7268.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2519, pruned_loss=0.05633, over 1443414.30 frames. ], batch size: 89, lr: 1.46e-02, grad_scale: 8.0 +2023-03-20 21:41:36,555 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29802.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:41:37,421 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 21:41:41,951 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.731e+02 2.521e+02 3.160e+02 3.849e+02 9.328e+02, threshold=6.319e+02, percent-clipped=4.0 +2023-03-20 21:41:56,883 INFO [train.py:901] (1/2) Epoch 11, batch 1600, loss[loss=0.1923, simple_loss=0.2624, pruned_loss=0.06112, over 7234.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2521, pruned_loss=0.05586, over 1444001.57 frames. ], batch size: 55, lr: 1.46e-02, grad_scale: 8.0 +2023-03-20 21:42:03,922 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-03-20 21:42:08,218 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 21:42:08,367 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29863.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:42:08,995 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 +2023-03-20 21:42:09,227 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 21:42:11,792 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 21:42:21,960 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 21:42:22,959 INFO [train.py:901] (1/2) Epoch 11, batch 1650, loss[loss=0.191, simple_loss=0.2627, pruned_loss=0.05961, over 7295.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2517, pruned_loss=0.05551, over 1442067.86 frames. ], batch size: 86, lr: 1.46e-02, grad_scale: 8.0 +2023-03-20 21:42:23,579 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29892.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:42:25,977 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 21:42:31,792 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29907.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:42:34,724 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.522e+02 2.457e+02 3.036e+02 3.652e+02 9.046e+02, threshold=6.071e+02, percent-clipped=4.0 +2023-03-20 21:42:34,767 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 21:42:48,513 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=29940.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:42:48,984 INFO [train.py:901] (1/2) Epoch 11, batch 1700, loss[loss=0.1586, simple_loss=0.2198, pruned_loss=0.04868, over 7004.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2513, pruned_loss=0.05523, over 1444955.19 frames. ], batch size: 35, lr: 1.46e-02, grad_scale: 8.0 +2023-03-20 21:42:51,541 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 21:42:55,247 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2438, 3.0318, 2.7656, 3.0693, 2.4239, 2.1930, 3.2036, 2.2952], + device='cuda:1'), covar=tensor([0.0187, 0.0268, 0.0205, 0.0208, 0.0271, 0.0361, 0.0233, 0.0690], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0277, 0.0233, 0.0281, 0.0302, 0.0295, 0.0273, 0.0297], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 21:42:55,566 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 21:43:06,933 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 21:43:15,695 INFO [train.py:901] (1/2) Epoch 11, batch 1750, loss[loss=0.1668, simple_loss=0.2399, pruned_loss=0.04686, over 7164.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2507, pruned_loss=0.05495, over 1442397.16 frames. ], batch size: 39, lr: 1.45e-02, grad_scale: 8.0 +2023-03-20 21:43:16,372 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9576, 2.0922, 1.9497, 3.0506, 1.3463, 3.2314, 1.2465, 2.9548], + device='cuda:1'), covar=tensor([0.0071, 0.0889, 0.1598, 0.0058, 0.4298, 0.0048, 0.0977, 0.0140], + device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0269, 0.0313, 0.0147, 0.0299, 0.0148, 0.0271, 0.0198], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 21:43:27,041 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.468e+02 2.815e+02 3.490e+02 6.653e+02, threshold=5.630e+02, percent-clipped=1.0 +2023-03-20 21:43:29,594 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30018.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 21:43:31,601 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3874, 3.5020, 3.2959, 3.4712, 3.2928, 3.3092, 3.6711, 3.6655], + device='cuda:1'), covar=tensor([0.0201, 0.0169, 0.0227, 0.0208, 0.0304, 0.0438, 0.0227, 0.0182], + device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0095, 0.0089, 0.0099, 0.0093, 0.0079, 0.0075, 0.0079], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:43:32,005 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 21:43:32,999 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 21:43:41,050 INFO [train.py:901] (1/2) Epoch 11, batch 1800, loss[loss=0.223, simple_loss=0.2885, pruned_loss=0.07876, over 6611.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2514, pruned_loss=0.05549, over 1440961.88 frames. ], batch size: 106, lr: 1.45e-02, grad_scale: 8.0 +2023-03-20 21:43:54,761 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 21:43:55,363 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4381, 2.0453, 2.4773, 1.4894, 1.5450, 1.5233, 1.7990, 1.7482], + device='cuda:1'), covar=tensor([0.0348, 0.0193, 0.0113, 0.0089, 0.0495, 0.0344, 0.0123, 0.0179], + device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0020, 0.0021, 0.0019, 0.0019, 0.0017, 0.0021, 0.0020], + device='cuda:1'), out_proj_covar=tensor([4.9732e-05, 4.8800e-05, 4.6956e-05, 4.0728e-05, 4.7670e-05, 4.3246e-05, + 4.8971e-05, 5.0514e-05], device='cuda:1') +2023-03-20 21:44:07,488 INFO [train.py:901] (1/2) Epoch 11, batch 1850, loss[loss=0.1583, simple_loss=0.2298, pruned_loss=0.04343, over 7179.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2521, pruned_loss=0.0561, over 1443050.22 frames. ], batch size: 39, lr: 1.45e-02, grad_scale: 8.0 +2023-03-20 21:44:08,547 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 21:44:17,446 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 21:44:18,414 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.849e+02 3.186e+02 3.684e+02 5.512e+02, threshold=6.372e+02, percent-clipped=0.0 +2023-03-20 21:44:23,624 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7708, 4.4158, 4.3646, 4.8494, 4.8042, 4.8798, 4.2192, 4.4755], + device='cuda:1'), covar=tensor([0.0799, 0.1978, 0.2076, 0.0911, 0.0707, 0.1096, 0.0775, 0.1008], + device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0282, 0.0236, 0.0223, 0.0178, 0.0289, 0.0164, 0.0202], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:44:24,723 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30125.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:44:33,215 INFO [train.py:901] (1/2) Epoch 11, batch 1900, loss[loss=0.1479, simple_loss=0.1948, pruned_loss=0.05056, over 5905.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2527, pruned_loss=0.0563, over 1443099.40 frames. ], batch size: 25, lr: 1.45e-02, grad_scale: 8.0 +2023-03-20 21:44:34,254 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 21:44:38,873 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2213, 3.7675, 3.9497, 3.8862, 3.5908, 3.7820, 4.1386, 3.6654], + device='cuda:1'), covar=tensor([0.0120, 0.0135, 0.0100, 0.0108, 0.0274, 0.0092, 0.0143, 0.0125], + device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0063, 0.0063, 0.0052, 0.0096, 0.0067, 0.0065, 0.0063], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:44:41,817 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30158.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:44:56,574 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30186.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:44:58,953 INFO [train.py:901] (1/2) Epoch 11, batch 1950, loss[loss=0.1764, simple_loss=0.2503, pruned_loss=0.05128, over 7258.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.252, pruned_loss=0.05614, over 1440487.58 frames. ], batch size: 52, lr: 1.45e-02, grad_scale: 8.0 +2023-03-20 21:44:59,974 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 21:45:07,601 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30207.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:45:08,669 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30209.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:45:10,570 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.824e+02 2.624e+02 3.249e+02 3.958e+02 8.410e+02, threshold=6.499e+02, percent-clipped=2.0 +2023-03-20 21:45:11,146 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 21:45:12,691 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=6.17 vs. limit=5.0 +2023-03-20 21:45:16,909 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 21:45:17,429 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 21:45:17,802 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-20 21:45:20,750 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 +2023-03-20 21:45:21,134 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1232, 2.3163, 2.2816, 2.9789, 1.3866, 3.4227, 1.2656, 2.8827], + device='cuda:1'), covar=tensor([0.0057, 0.0875, 0.1615, 0.0057, 0.4086, 0.0052, 0.1098, 0.0089], + device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0264, 0.0313, 0.0146, 0.0297, 0.0147, 0.0272, 0.0197], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 21:45:23,192 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8683, 2.1074, 1.8229, 2.8905, 2.6613, 2.2865, 2.2406, 1.7811], + device='cuda:1'), covar=tensor([0.1667, 0.0740, 0.2138, 0.0423, 0.0108, 0.0045, 0.0110, 0.0088], + device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0222, 0.0272, 0.0248, 0.0126, 0.0118, 0.0142, 0.0158], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:45:25,566 INFO [train.py:901] (1/2) Epoch 11, batch 2000, loss[loss=0.1847, simple_loss=0.2502, pruned_loss=0.05962, over 7242.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2517, pruned_loss=0.0563, over 1438181.21 frames. ], batch size: 89, lr: 1.45e-02, grad_scale: 8.0 +2023-03-20 21:45:30,357 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4586, 3.5865, 2.4870, 4.0977, 3.1878, 3.6857, 2.3940, 2.1844], + device='cuda:1'), covar=tensor([0.0195, 0.0473, 0.1327, 0.0239, 0.0297, 0.0281, 0.1505, 0.1320], + device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0214, 0.0306, 0.0203, 0.0229, 0.0219, 0.0274, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-20 21:45:33,273 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=30255.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:45:33,739 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 21:45:35,848 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0793, 4.0144, 3.5077, 3.1504, 3.3331, 1.9455, 1.6193, 3.9495], + device='cuda:1'), covar=tensor([0.0024, 0.0051, 0.0085, 0.0100, 0.0100, 0.0579, 0.0664, 0.0050], + device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0060, 0.0079, 0.0069, 0.0084, 0.0104, 0.0109, 0.0071], + device='cuda:1'), out_proj_covar=tensor([8.0654e-05, 8.9368e-05, 1.0739e-04, 1.0196e-04, 1.1512e-04, 1.4569e-04, + 1.5139e-04, 9.8449e-05], device='cuda:1') +2023-03-20 21:45:40,894 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30270.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:45:44,344 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 21:45:51,051 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-03-20 21:45:51,248 INFO [train.py:901] (1/2) Epoch 11, batch 2050, loss[loss=0.203, simple_loss=0.2703, pruned_loss=0.06786, over 7273.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2517, pruned_loss=0.05628, over 1440814.42 frames. ], batch size: 52, lr: 1.45e-02, grad_scale: 8.0 +2023-03-20 21:45:52,760 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 21:46:03,019 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.450e+02 2.835e+02 3.331e+02 9.536e+02, threshold=5.670e+02, percent-clipped=1.0 +2023-03-20 21:46:05,770 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30318.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 21:46:06,785 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8940, 3.7290, 3.4797, 3.4952, 2.9201, 3.6028, 3.6590, 3.5388], + device='cuda:1'), covar=tensor([0.0159, 0.0169, 0.0205, 0.0200, 0.0503, 0.0151, 0.0251, 0.0171], + device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0063, 0.0064, 0.0052, 0.0099, 0.0068, 0.0065, 0.0064], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:46:17,839 INFO [train.py:901] (1/2) Epoch 11, batch 2100, loss[loss=0.1668, simple_loss=0.2423, pruned_loss=0.04565, over 7270.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2513, pruned_loss=0.05608, over 1441288.37 frames. ], batch size: 47, lr: 1.45e-02, grad_scale: 8.0 +2023-03-20 21:46:26,945 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 21:46:27,612 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30360.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:46:29,577 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9500, 3.4676, 3.7771, 3.8115, 3.7237, 3.8226, 3.9083, 3.5522], + device='cuda:1'), covar=tensor([0.0041, 0.0113, 0.0048, 0.0048, 0.0047, 0.0051, 0.0064, 0.0072], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0041, 0.0037, 0.0035, 0.0035, 0.0036, 0.0042, 0.0044], + device='cuda:1'), out_proj_covar=tensor([7.8760e-05, 1.2102e-04, 1.0782e-04, 9.1449e-05, 8.8954e-05, 9.6273e-05, + 1.2263e-04, 1.1843e-04], device='cuda:1') +2023-03-20 21:46:30,485 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 21:46:30,531 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=30366.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:46:42,970 INFO [train.py:901] (1/2) Epoch 11, batch 2150, loss[loss=0.21, simple_loss=0.2795, pruned_loss=0.07028, over 7255.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.251, pruned_loss=0.05591, over 1441632.81 frames. ], batch size: 64, lr: 1.45e-02, grad_scale: 8.0 +2023-03-20 21:46:54,911 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.318e+02 2.491e+02 3.235e+02 3.688e+02 7.373e+02, threshold=6.470e+02, percent-clipped=3.0 +2023-03-20 21:46:59,057 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30421.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:46:59,783 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.10 vs. limit=2.0 +2023-03-20 21:47:09,683 INFO [train.py:901] (1/2) Epoch 11, batch 2200, loss[loss=0.1833, simple_loss=0.2534, pruned_loss=0.05661, over 7364.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2501, pruned_loss=0.05516, over 1443195.83 frames. ], batch size: 63, lr: 1.44e-02, grad_scale: 8.0 +2023-03-20 21:47:16,862 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 21:47:18,506 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30458.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:47:23,959 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4478, 4.9098, 4.9084, 4.8462, 4.7583, 4.5051, 4.9340, 4.7844], + device='cuda:1'), covar=tensor([0.0358, 0.0326, 0.0302, 0.0370, 0.0286, 0.0297, 0.0283, 0.0372], + device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0174, 0.0125, 0.0129, 0.0113, 0.0164, 0.0140, 0.0108], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:47:29,241 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0132, 0.8683, 1.0876, 1.5596, 1.2222, 1.4418, 1.1106, 1.0715], + device='cuda:1'), covar=tensor([0.1566, 0.2418, 0.1036, 0.0470, 0.2044, 0.1667, 0.1096, 0.2179], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0039, 0.0029, 0.0030, 0.0036, 0.0033, 0.0039, 0.0033], + device='cuda:1'), out_proj_covar=tensor([7.8982e-05, 9.2820e-05, 6.8863e-05, 7.0106e-05, 8.2092e-05, 7.9813e-05, + 8.9213e-05, 8.1077e-05], device='cuda:1') +2023-03-20 21:47:30,626 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30481.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:47:35,660 INFO [train.py:901] (1/2) Epoch 11, batch 2250, loss[loss=0.1882, simple_loss=0.2614, pruned_loss=0.05756, over 7270.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2506, pruned_loss=0.05556, over 1441344.58 frames. ], batch size: 77, lr: 1.44e-02, grad_scale: 8.0 +2023-03-20 21:47:43,317 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=30506.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:47:47,338 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.268e+02 2.464e+02 3.038e+02 3.870e+02 9.111e+02, threshold=6.075e+02, percent-clipped=2.0 +2023-03-20 21:47:50,876 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 21:47:50,887 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 21:47:53,958 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0847, 4.5915, 4.6518, 4.6179, 4.5066, 4.1199, 4.6632, 4.5082], + device='cuda:1'), covar=tensor([0.0480, 0.0374, 0.0334, 0.0333, 0.0319, 0.0363, 0.0281, 0.0421], + device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0173, 0.0126, 0.0127, 0.0112, 0.0163, 0.0139, 0.0108], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:48:01,420 INFO [train.py:901] (1/2) Epoch 11, batch 2300, loss[loss=0.2281, simple_loss=0.2923, pruned_loss=0.08202, over 6725.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.251, pruned_loss=0.05559, over 1441890.76 frames. ], batch size: 107, lr: 1.44e-02, grad_scale: 8.0 +2023-03-20 21:48:02,211 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.64 vs. limit=5.0 +2023-03-20 21:48:02,917 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 21:48:09,125 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5815, 3.3908, 3.1294, 3.2345, 2.6952, 2.4378, 3.6740, 2.5040], + device='cuda:1'), covar=tensor([0.0184, 0.0200, 0.0216, 0.0198, 0.0281, 0.0451, 0.0256, 0.0654], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0278, 0.0236, 0.0288, 0.0303, 0.0300, 0.0274, 0.0293], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 21:48:14,266 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30565.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:48:15,874 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3260, 2.6153, 2.0950, 2.5950, 2.6824, 2.1682, 2.7345, 2.5343], + device='cuda:1'), covar=tensor([0.1184, 0.0648, 0.1189, 0.1357, 0.1055, 0.1335, 0.1197, 0.1041], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0035, 0.0042, 0.0037, 0.0039, 0.0038, 0.0035, 0.0035], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 21:48:27,196 INFO [train.py:901] (1/2) Epoch 11, batch 2350, loss[loss=0.1799, simple_loss=0.2542, pruned_loss=0.05274, over 7260.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2514, pruned_loss=0.05603, over 1442646.68 frames. ], batch size: 89, lr: 1.44e-02, grad_scale: 8.0 +2023-03-20 21:48:30,241 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-20 21:48:31,487 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9736, 3.9933, 3.8839, 4.0833, 3.7746, 4.0511, 4.4204, 4.4686], + device='cuda:1'), covar=tensor([0.0159, 0.0162, 0.0166, 0.0155, 0.0325, 0.0206, 0.0174, 0.0128], + device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0094, 0.0087, 0.0098, 0.0094, 0.0078, 0.0074, 0.0076], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:48:31,743 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 +2023-03-20 21:48:39,219 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.571e+02 3.025e+02 3.938e+02 7.428e+02, threshold=6.050e+02, percent-clipped=3.0 +2023-03-20 21:48:50,826 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 21:48:53,295 INFO [train.py:901] (1/2) Epoch 11, batch 2400, loss[loss=0.1955, simple_loss=0.2676, pruned_loss=0.0617, over 7354.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.251, pruned_loss=0.05537, over 1444978.64 frames. ], batch size: 73, lr: 1.44e-02, grad_scale: 8.0 +2023-03-20 21:48:57,955 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 21:48:58,609 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0506, 0.8946, 1.1838, 1.5152, 1.1913, 1.5579, 1.3252, 1.1741], + device='cuda:1'), covar=tensor([0.1408, 0.1673, 0.0693, 0.0586, 0.1473, 0.1159, 0.0978, 0.1761], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0042, 0.0030, 0.0031, 0.0038, 0.0035, 0.0040, 0.0034], + device='cuda:1'), out_proj_covar=tensor([8.3198e-05, 9.7491e-05, 7.0746e-05, 7.2905e-05, 8.6037e-05, 8.4129e-05, + 9.2467e-05, 8.3794e-05], device='cuda:1') +2023-03-20 21:49:01,145 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30655.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:49:03,315 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-03-20 21:49:08,238 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 21:49:11,254 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 21:49:19,891 INFO [train.py:901] (1/2) Epoch 11, batch 2450, loss[loss=0.17, simple_loss=0.2395, pruned_loss=0.05024, over 7201.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2505, pruned_loss=0.05513, over 1442838.23 frames. ], batch size: 50, lr: 1.44e-02, grad_scale: 8.0 +2023-03-20 21:49:25,630 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3998, 2.9586, 3.2864, 3.1738, 3.3058, 3.3635, 3.1161, 3.2807], + device='cuda:1'), covar=tensor([0.0028, 0.0085, 0.0038, 0.0047, 0.0035, 0.0032, 0.0075, 0.0052], + device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0041, 0.0037, 0.0035, 0.0035, 0.0037, 0.0042, 0.0044], + device='cuda:1'), out_proj_covar=tensor([8.0309e-05, 1.1884e-04, 1.0454e-04, 9.2524e-05, 8.9655e-05, 9.8659e-05, + 1.2252e-04, 1.1785e-04], device='cuda:1') +2023-03-20 21:49:31,056 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.414e+02 2.901e+02 3.572e+02 6.150e+02, threshold=5.802e+02, percent-clipped=2.0 +2023-03-20 21:49:32,673 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30716.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:49:32,730 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30716.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 21:49:38,131 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 21:49:45,814 INFO [train.py:901] (1/2) Epoch 11, batch 2500, loss[loss=0.174, simple_loss=0.2491, pruned_loss=0.04944, over 7261.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2508, pruned_loss=0.05523, over 1442091.87 frames. ], batch size: 77, lr: 1.44e-02, grad_scale: 8.0 +2023-03-20 21:49:59,714 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5606, 3.0840, 3.2301, 3.3314, 3.5708, 3.5096, 3.2744, 3.2959], + device='cuda:1'), covar=tensor([0.0029, 0.0088, 0.0049, 0.0061, 0.0031, 0.0036, 0.0074, 0.0062], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0040, 0.0037, 0.0035, 0.0035, 0.0037, 0.0042, 0.0044], + device='cuda:1'), out_proj_covar=tensor([7.9277e-05, 1.1815e-04, 1.0464e-04, 9.2603e-05, 8.8954e-05, 9.6662e-05, + 1.2228e-04, 1.1802e-04], device='cuda:1') +2023-03-20 21:50:03,718 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 21:50:06,816 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30781.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:50:11,658 INFO [train.py:901] (1/2) Epoch 11, batch 2550, loss[loss=0.2313, simple_loss=0.2929, pruned_loss=0.0849, over 6632.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2513, pruned_loss=0.0558, over 1441084.31 frames. ], batch size: 106, lr: 1.44e-02, grad_scale: 8.0 +2023-03-20 21:50:13,295 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2843, 1.5996, 1.5152, 1.2354, 1.7050, 1.3375, 1.1805, 1.4084], + device='cuda:1'), covar=tensor([0.0177, 0.0326, 0.0171, 0.0120, 0.0246, 0.0274, 0.0221, 0.0207], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0021, 0.0022, 0.0020, 0.0021, 0.0020, 0.0022, 0.0022], + device='cuda:1'), out_proj_covar=tensor([5.2198e-05, 5.1545e-05, 5.0835e-05, 4.4275e-05, 5.2207e-05, 4.7621e-05, + 5.2159e-05, 5.4917e-05], device='cuda:1') +2023-03-20 21:50:22,946 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.481e+02 2.521e+02 2.930e+02 3.805e+02 7.129e+02, threshold=5.860e+02, percent-clipped=4.0 +2023-03-20 21:50:23,113 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9291, 2.1975, 1.9194, 2.0874, 2.2964, 1.8875, 2.3874, 2.2128], + device='cuda:1'), covar=tensor([0.0838, 0.0495, 0.0620, 0.0554, 0.0542, 0.0564, 0.0407, 0.0457], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0036, 0.0042, 0.0038, 0.0039, 0.0038, 0.0036, 0.0034], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 21:50:31,752 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=30829.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:50:37,736 INFO [train.py:901] (1/2) Epoch 11, batch 2600, loss[loss=0.1569, simple_loss=0.2332, pruned_loss=0.0403, over 7322.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2516, pruned_loss=0.05587, over 1442117.19 frames. ], batch size: 75, lr: 1.43e-02, grad_scale: 8.0 +2023-03-20 21:50:49,934 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30865.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:50:51,044 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3494, 2.9645, 2.9602, 2.9239, 2.5901, 2.4626, 3.2046, 2.3185], + device='cuda:1'), covar=tensor([0.0159, 0.0180, 0.0241, 0.0214, 0.0345, 0.0392, 0.0247, 0.0697], + device='cuda:1'), in_proj_covar=tensor([0.0266, 0.0274, 0.0237, 0.0286, 0.0301, 0.0292, 0.0272, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 21:50:59,006 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3609, 1.3990, 1.3512, 1.2461, 1.2633, 1.2036, 1.0214, 1.0723], + device='cuda:1'), covar=tensor([0.0084, 0.0089, 0.0109, 0.0096, 0.0145, 0.0078, 0.0253, 0.0117], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0018, 0.0019, 0.0019, 0.0021, 0.0019, 0.0020, 0.0025], + device='cuda:1'), out_proj_covar=tensor([2.4719e-05, 2.1198e-05, 2.3434e-05, 2.0872e-05, 2.5641e-05, 2.0964e-05, + 2.3589e-05, 3.0630e-05], device='cuda:1') +2023-03-20 21:51:03,255 INFO [train.py:901] (1/2) Epoch 11, batch 2650, loss[loss=0.1679, simple_loss=0.2408, pruned_loss=0.04755, over 7336.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2511, pruned_loss=0.05534, over 1444494.03 frames. ], batch size: 44, lr: 1.43e-02, grad_scale: 8.0 +2023-03-20 21:51:09,396 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30903.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:51:11,392 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2608, 1.5868, 1.2542, 1.1801, 1.2230, 1.1841, 1.0055, 1.0476], + device='cuda:1'), covar=tensor([0.0088, 0.0065, 0.0292, 0.0066, 0.0104, 0.0072, 0.0150, 0.0148], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0018, 0.0019, 0.0019, 0.0021, 0.0019, 0.0020, 0.0025], + device='cuda:1'), out_proj_covar=tensor([2.4895e-05, 2.1371e-05, 2.3895e-05, 2.1009e-05, 2.5890e-05, 2.0933e-05, + 2.3974e-05, 3.0920e-05], device='cuda:1') +2023-03-20 21:51:14,264 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 2.453e+02 2.842e+02 3.553e+02 6.219e+02, threshold=5.685e+02, percent-clipped=2.0 +2023-03-20 21:51:14,317 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=30913.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:51:20,307 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30925.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:51:24,814 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30934.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:51:28,158 INFO [train.py:901] (1/2) Epoch 11, batch 2700, loss[loss=0.1794, simple_loss=0.2531, pruned_loss=0.05285, over 7278.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2503, pruned_loss=0.0551, over 1442669.40 frames. ], batch size: 77, lr: 1.43e-02, grad_scale: 8.0 +2023-03-20 21:51:39,721 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30964.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:51:50,666 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30986.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:51:53,058 INFO [train.py:901] (1/2) Epoch 11, batch 2750, loss[loss=0.1821, simple_loss=0.2543, pruned_loss=0.0549, over 7312.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2495, pruned_loss=0.05507, over 1441999.96 frames. ], batch size: 59, lr: 1.43e-02, grad_scale: 8.0 +2023-03-20 21:51:55,170 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30995.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:51:56,136 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1683, 1.3042, 1.2978, 2.0160, 1.4706, 1.6130, 1.5255, 1.4575], + device='cuda:1'), covar=tensor([0.2611, 0.2880, 0.0532, 0.0549, 0.2157, 0.1097, 0.0931, 0.2163], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0042, 0.0032, 0.0032, 0.0040, 0.0038, 0.0042, 0.0036], + device='cuda:1'), out_proj_covar=tensor([8.6964e-05, 9.9170e-05, 7.5230e-05, 7.6122e-05, 8.9661e-05, 8.9118e-05, + 9.6675e-05, 8.7775e-05], device='cuda:1') +2023-03-20 21:52:03,410 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31011.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 21:52:03,446 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31011.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 21:52:04,279 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.554e+02 2.469e+02 3.015e+02 3.883e+02 6.735e+02, threshold=6.029e+02, percent-clipped=4.0 +2023-03-20 21:52:05,829 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31016.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:52:12,842 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0526, 2.2868, 1.6659, 2.8437, 2.0709, 2.6175, 1.8642, 1.9204], + device='cuda:1'), covar=tensor([0.1510, 0.0634, 0.2360, 0.0518, 0.0096, 0.0054, 0.0095, 0.0101], + device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0221, 0.0275, 0.0245, 0.0129, 0.0119, 0.0141, 0.0153], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:52:18,065 INFO [train.py:901] (1/2) Epoch 11, batch 2800, loss[loss=0.2342, simple_loss=0.3003, pruned_loss=0.08411, over 6783.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2498, pruned_loss=0.05514, over 1443643.61 frames. ], batch size: 107, lr: 1.43e-02, grad_scale: 8.0 +2023-03-20 21:52:44,653 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 21:52:44,707 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 21:52:44,732 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 21:52:44,748 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 21:52:44,921 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 21:52:44,937 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 21:52:44,947 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 21:52:45,336 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 21:52:45,414 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 21:52:45,428 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 21:52:45,445 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 21:52:45,601 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 21:52:45,606 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 21:52:45,619 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 21:52:45,753 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 21:52:45,756 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 21:52:45,815 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 21:52:53,165 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=31064.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:52:53,617 INFO [train.py:901] (1/2) Epoch 12, batch 0, loss[loss=0.1588, simple_loss=0.2301, pruned_loss=0.0437, over 7200.00 frames. ], tot_loss[loss=0.1588, simple_loss=0.2301, pruned_loss=0.0437, over 7200.00 frames. ], batch size: 39, lr: 1.38e-02, grad_scale: 8.0 +2023-03-20 21:52:53,617 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 21:52:58,003 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0921, 3.8584, 3.5381, 3.3832, 3.6497, 2.4357, 1.6474, 3.9326], + device='cuda:1'), covar=tensor([0.0024, 0.0085, 0.0046, 0.0065, 0.0050, 0.0415, 0.0589, 0.0042], + device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0060, 0.0078, 0.0070, 0.0083, 0.0104, 0.0108, 0.0071], + device='cuda:1'), out_proj_covar=tensor([8.1462e-05, 8.8969e-05, 1.0513e-04, 1.0114e-04, 1.1331e-04, 1.4381e-04, + 1.4938e-04, 9.8415e-05], device='cuda:1') +2023-03-20 21:52:59,695 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6488, 4.9853, 4.8954, 4.9737, 4.6982, 4.6012, 4.9710, 4.7526], + device='cuda:1'), covar=tensor([0.0404, 0.0403, 0.0531, 0.0470, 0.0348, 0.0277, 0.0355, 0.0536], + device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0174, 0.0127, 0.0127, 0.0112, 0.0162, 0.0141, 0.0107], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:53:20,415 INFO [train.py:935] (1/2) Epoch 12, validation: loss=0.1724, simple_loss=0.26, pruned_loss=0.04238, over 1622729.00 frames. +2023-03-20 21:53:20,416 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-20 21:53:24,073 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31072.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 21:53:27,954 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 21:53:37,907 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 21:53:44,591 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 21:53:45,086 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.395e+02 2.570e+02 3.180e+02 3.808e+02 7.988e+02, threshold=6.360e+02, percent-clipped=6.0 +2023-03-20 21:53:46,117 INFO [train.py:901] (1/2) Epoch 12, batch 50, loss[loss=0.194, simple_loss=0.2611, pruned_loss=0.06343, over 7275.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2513, pruned_loss=0.05427, over 323907.67 frames. ], batch size: 66, lr: 1.38e-02, grad_scale: 8.0 +2023-03-20 21:53:47,128 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 21:53:49,620 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 21:53:51,270 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8780, 1.9942, 1.6806, 2.6639, 1.9587, 2.7118, 2.1363, 1.7871], + device='cuda:1'), covar=tensor([0.1777, 0.0826, 0.2829, 0.0648, 0.0080, 0.0068, 0.0116, 0.0092], + device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0223, 0.0277, 0.0248, 0.0129, 0.0120, 0.0141, 0.0155], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:54:05,383 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 21:54:05,874 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 21:54:11,854 INFO [train.py:901] (1/2) Epoch 12, batch 100, loss[loss=0.1718, simple_loss=0.2486, pruned_loss=0.04747, over 7289.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2509, pruned_loss=0.05374, over 574936.36 frames. ], batch size: 77, lr: 1.38e-02, grad_scale: 8.0 +2023-03-20 21:54:15,033 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1764, 1.4314, 1.1431, 1.2312, 1.1242, 1.1362, 1.1198, 0.9091], + device='cuda:1'), covar=tensor([0.0317, 0.0148, 0.0229, 0.0083, 0.0180, 0.0125, 0.0251, 0.0147], + device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0019, 0.0019, 0.0018, 0.0020, 0.0019, 0.0020, 0.0025], + device='cuda:1'), out_proj_covar=tensor([2.4469e-05, 2.2271e-05, 2.3785e-05, 2.0369e-05, 2.5045e-05, 2.0985e-05, + 2.3265e-05, 3.0940e-05], device='cuda:1') +2023-03-20 21:54:16,536 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4815, 3.3378, 3.0506, 3.1937, 2.7628, 2.5208, 3.4786, 2.3902], + device='cuda:1'), covar=tensor([0.0187, 0.0182, 0.0234, 0.0184, 0.0272, 0.0399, 0.0274, 0.0735], + device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0275, 0.0243, 0.0289, 0.0300, 0.0296, 0.0276, 0.0294], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 21:54:32,054 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1837, 1.6419, 1.8863, 1.3201, 1.6773, 1.6417, 1.4633, 1.4281], + device='cuda:1'), covar=tensor([0.0328, 0.0339, 0.0072, 0.0098, 0.0371, 0.0229, 0.0194, 0.0155], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0020, 0.0020, 0.0019, 0.0019, 0.0018, 0.0020, 0.0020], + device='cuda:1'), out_proj_covar=tensor([5.0573e-05, 4.8222e-05, 4.5725e-05, 4.1140e-05, 4.7817e-05, 4.3577e-05, + 4.7190e-05, 5.1140e-05], device='cuda:1') +2023-03-20 21:54:36,334 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-03-20 21:54:36,935 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.411e+02 2.828e+02 3.722e+02 7.099e+02, threshold=5.657e+02, percent-clipped=3.0 +2023-03-20 21:54:37,966 INFO [train.py:901] (1/2) Epoch 12, batch 150, loss[loss=0.1652, simple_loss=0.242, pruned_loss=0.04418, over 7223.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2506, pruned_loss=0.0539, over 766796.11 frames. ], batch size: 45, lr: 1.38e-02, grad_scale: 8.0 +2023-03-20 21:54:44,185 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7208, 4.4655, 4.3882, 4.9777, 4.9021, 4.9494, 4.2786, 4.4451], + device='cuda:1'), covar=tensor([0.0770, 0.2095, 0.1675, 0.0843, 0.0609, 0.0978, 0.0785, 0.0991], + device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0280, 0.0230, 0.0220, 0.0173, 0.0283, 0.0159, 0.0199], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:55:00,543 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31259.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:55:03,406 INFO [train.py:901] (1/2) Epoch 12, batch 200, loss[loss=0.1944, simple_loss=0.2631, pruned_loss=0.06282, over 7308.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.249, pruned_loss=0.05306, over 917138.21 frames. ], batch size: 59, lr: 1.37e-02, grad_scale: 8.0 +2023-03-20 21:55:04,991 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 21:55:09,965 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 21:55:12,050 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31281.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:55:16,372 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 21:55:16,434 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31290.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:55:26,935 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31311.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:55:27,809 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.519e+02 2.425e+02 2.946e+02 3.483e+02 6.440e+02, threshold=5.892e+02, percent-clipped=1.0 +2023-03-20 21:55:28,880 INFO [train.py:901] (1/2) Epoch 12, batch 250, loss[loss=0.1678, simple_loss=0.214, pruned_loss=0.06078, over 6346.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2487, pruned_loss=0.05314, over 1032968.57 frames. ], batch size: 27, lr: 1.37e-02, grad_scale: 8.0 +2023-03-20 21:55:29,997 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 21:55:30,233 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 +2023-03-20 21:55:31,550 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7487, 4.4534, 4.4857, 4.9763, 4.9193, 4.9541, 4.2862, 4.4813], + device='cuda:1'), covar=tensor([0.0930, 0.2405, 0.2247, 0.1006, 0.0861, 0.1130, 0.0719, 0.1078], + device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0289, 0.0237, 0.0228, 0.0180, 0.0293, 0.0162, 0.0205], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:55:50,524 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 21:55:51,521 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=31359.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:55:54,415 INFO [train.py:901] (1/2) Epoch 12, batch 300, loss[loss=0.1816, simple_loss=0.2441, pruned_loss=0.05957, over 7230.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2473, pruned_loss=0.05281, over 1122142.16 frames. ], batch size: 55, lr: 1.37e-02, grad_scale: 8.0 +2023-03-20 21:55:55,451 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31367.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 21:56:00,014 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 21:56:18,389 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31409.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:56:20,278 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.464e+02 2.963e+02 3.732e+02 1.202e+03, threshold=5.926e+02, percent-clipped=2.0 +2023-03-20 21:56:21,317 INFO [train.py:901] (1/2) Epoch 12, batch 350, loss[loss=0.147, simple_loss=0.206, pruned_loss=0.044, over 5839.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2494, pruned_loss=0.05433, over 1192001.57 frames. ], batch size: 25, lr: 1.37e-02, grad_scale: 8.0 +2023-03-20 21:56:35,611 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 21:56:47,121 INFO [train.py:901] (1/2) Epoch 12, batch 400, loss[loss=0.1751, simple_loss=0.248, pruned_loss=0.0511, over 7258.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2486, pruned_loss=0.05344, over 1247715.25 frames. ], batch size: 64, lr: 1.37e-02, grad_scale: 8.0 +2023-03-20 21:56:49,776 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31470.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:57:00,314 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31491.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 21:57:11,814 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 2.122e+02 2.465e+02 3.114e+02 1.072e+03, threshold=4.930e+02, percent-clipped=4.0 +2023-03-20 21:57:12,848 INFO [train.py:901] (1/2) Epoch 12, batch 450, loss[loss=0.1926, simple_loss=0.2621, pruned_loss=0.0616, over 7343.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2483, pruned_loss=0.05289, over 1293671.91 frames. ], batch size: 61, lr: 1.37e-02, grad_scale: 8.0 +2023-03-20 21:57:16,374 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 21:57:16,892 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 21:57:32,075 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31552.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 21:57:35,453 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31559.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:57:38,344 INFO [train.py:901] (1/2) Epoch 12, batch 500, loss[loss=0.1836, simple_loss=0.2591, pruned_loss=0.05411, over 7201.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2488, pruned_loss=0.053, over 1328503.86 frames. ], batch size: 99, lr: 1.37e-02, grad_scale: 16.0 +2023-03-20 21:57:41,198 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.02 vs. limit=2.0 +2023-03-20 21:57:47,061 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31581.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:57:50,034 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 21:57:51,503 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 21:57:51,623 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31590.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:57:52,530 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 21:57:54,557 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 21:57:58,839 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 21:58:00,364 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=31607.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:58:00,732 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 +2023-03-20 21:58:03,281 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.477e+02 2.904e+02 3.674e+02 7.030e+02, threshold=5.807e+02, percent-clipped=4.0 +2023-03-20 21:58:04,290 INFO [train.py:901] (1/2) Epoch 12, batch 550, loss[loss=0.1863, simple_loss=0.2578, pruned_loss=0.05734, over 7277.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2496, pruned_loss=0.05362, over 1355295.83 frames. ], batch size: 68, lr: 1.37e-02, grad_scale: 16.0 +2023-03-20 21:58:04,610 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-20 21:58:10,831 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 21:58:11,862 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=31629.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:58:16,375 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=31638.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:58:18,825 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 21:58:22,342 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 21:58:22,459 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4363, 2.8034, 2.1644, 2.9763, 2.6808, 2.1918, 2.7919, 2.5957], + device='cuda:1'), covar=tensor([0.0533, 0.0520, 0.2044, 0.0760, 0.1351, 0.0859, 0.0871, 0.0937], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0038, 0.0045, 0.0040, 0.0041, 0.0039, 0.0039, 0.0036], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 21:58:23,406 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3208, 1.4199, 1.6824, 1.1563, 1.6796, 1.8479, 1.4091, 1.3875], + device='cuda:1'), covar=tensor([0.0132, 0.0296, 0.0121, 0.0075, 0.0319, 0.0138, 0.0140, 0.0163], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0020, 0.0021, 0.0019, 0.0019, 0.0018, 0.0019, 0.0020], + device='cuda:1'), out_proj_covar=tensor([5.0918e-05, 4.9353e-05, 4.7682e-05, 4.2296e-05, 4.7416e-05, 4.4030e-05, + 4.6657e-05, 5.1204e-05], device='cuda:1') +2023-03-20 21:58:29,250 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 21:58:29,744 INFO [train.py:901] (1/2) Epoch 12, batch 600, loss[loss=0.1856, simple_loss=0.2574, pruned_loss=0.05691, over 7295.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.25, pruned_loss=0.05411, over 1376195.70 frames. ], batch size: 66, lr: 1.37e-02, grad_scale: 16.0 +2023-03-20 21:58:31,455 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31667.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 21:58:32,025 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0937, 1.0992, 1.0513, 1.5539, 1.4957, 1.5225, 1.2158, 1.2320], + device='cuda:1'), covar=tensor([0.0772, 0.1339, 0.0882, 0.0427, 0.0669, 0.0650, 0.0576, 0.1357], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0041, 0.0032, 0.0031, 0.0038, 0.0037, 0.0041, 0.0036], + device='cuda:1'), out_proj_covar=tensor([8.5930e-05, 9.9120e-05, 7.6249e-05, 7.5939e-05, 8.8582e-05, 8.8874e-05, + 9.6259e-05, 8.8628e-05], device='cuda:1') +2023-03-20 21:58:36,987 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7880, 3.3480, 3.5332, 3.4387, 3.2475, 3.3225, 3.6994, 3.4503], + device='cuda:1'), covar=tensor([0.0105, 0.0171, 0.0114, 0.0148, 0.0365, 0.0118, 0.0142, 0.0119], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0065, 0.0066, 0.0054, 0.0103, 0.0070, 0.0069, 0.0066], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 21:58:45,359 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 21:58:49,939 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31704.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:58:53,256 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 +2023-03-20 21:58:54,861 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.764e+02 2.406e+02 2.959e+02 3.734e+02 7.911e+02, threshold=5.917e+02, percent-clipped=4.0 +2023-03-20 21:58:55,405 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 21:58:55,888 INFO [train.py:901] (1/2) Epoch 12, batch 650, loss[loss=0.176, simple_loss=0.2455, pruned_loss=0.05328, over 7324.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2488, pruned_loss=0.05365, over 1390176.55 frames. ], batch size: 59, lr: 1.37e-02, grad_scale: 16.0 +2023-03-20 21:58:55,941 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=31715.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 21:59:03,683 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1031, 3.2747, 1.9587, 3.5970, 2.4392, 2.9429, 1.5835, 2.0199], + device='cuda:1'), covar=tensor([0.0186, 0.0545, 0.1543, 0.0297, 0.0272, 0.0352, 0.2146, 0.1204], + device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0216, 0.0302, 0.0208, 0.0232, 0.0217, 0.0271, 0.0286], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 21:59:12,135 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 21:59:22,005 INFO [train.py:901] (1/2) Epoch 12, batch 700, loss[loss=0.2076, simple_loss=0.2742, pruned_loss=0.07049, over 6699.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2473, pruned_loss=0.05336, over 1397428.34 frames. ], batch size: 106, lr: 1.36e-02, grad_scale: 16.0 +2023-03-20 21:59:22,070 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:59:22,158 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:59:22,512 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 21:59:33,743 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31788.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 21:59:46,998 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.502e+02 2.229e+02 2.839e+02 3.592e+02 6.099e+02, threshold=5.677e+02, percent-clipped=2.0 +2023-03-20 21:59:47,052 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 21:59:47,533 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 21:59:48,025 INFO [train.py:901] (1/2) Epoch 12, batch 750, loss[loss=0.1736, simple_loss=0.2513, pruned_loss=0.04791, over 7369.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2478, pruned_loss=0.05342, over 1407050.76 frames. ], batch size: 63, lr: 1.36e-02, grad_scale: 16.0 +2023-03-20 21:59:52,212 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31823.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:00:02,261 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 22:00:04,924 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31847.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 22:00:05,968 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31849.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:00:06,843 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 22:00:12,890 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 22:00:13,835 INFO [train.py:901] (1/2) Epoch 12, batch 800, loss[loss=0.1855, simple_loss=0.2613, pruned_loss=0.05487, over 7259.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.248, pruned_loss=0.05328, over 1415295.11 frames. ], batch size: 52, lr: 1.36e-02, grad_scale: 16.0 +2023-03-20 22:00:13,989 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1719, 2.4850, 2.1381, 2.5022, 2.4536, 2.2038, 2.5205, 2.4135], + device='cuda:1'), covar=tensor([0.0738, 0.0594, 0.0646, 0.0435, 0.0398, 0.0497, 0.0527, 0.0871], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0037, 0.0043, 0.0039, 0.0040, 0.0038, 0.0040, 0.0035], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 22:00:14,242 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.90 vs. limit=2.0 +2023-03-20 22:00:14,359 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 22:00:17,430 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7012, 4.3562, 4.2959, 4.7407, 4.7545, 4.7573, 3.9465, 4.2891], + device='cuda:1'), covar=tensor([0.0688, 0.1892, 0.1946, 0.0959, 0.0710, 0.1141, 0.0827, 0.0979], + device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0275, 0.0231, 0.0217, 0.0165, 0.0280, 0.0156, 0.0196], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:00:20,542 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31878.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:00:23,017 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5381, 4.1982, 4.0933, 4.6330, 4.6077, 4.5904, 4.0751, 4.0828], + device='cuda:1'), covar=tensor([0.0790, 0.2230, 0.2497, 0.1100, 0.0768, 0.1384, 0.0716, 0.1077], + device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0280, 0.0237, 0.0221, 0.0168, 0.0285, 0.0158, 0.0200], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:00:23,610 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31884.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:00:25,500 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 22:00:33,091 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7603, 2.0699, 1.6693, 2.6944, 2.3727, 2.4737, 2.0282, 2.1144], + device='cuda:1'), covar=tensor([0.1587, 0.0651, 0.2364, 0.0530, 0.0088, 0.0061, 0.0152, 0.0168], + device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0220, 0.0272, 0.0247, 0.0128, 0.0120, 0.0144, 0.0156], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:00:38,419 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.738e+02 2.334e+02 2.815e+02 3.523e+02 9.191e+02, threshold=5.630e+02, percent-clipped=3.0 +2023-03-20 22:00:39,480 INFO [train.py:901] (1/2) Epoch 12, batch 850, loss[loss=0.1769, simple_loss=0.2503, pruned_loss=0.05173, over 7210.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2485, pruned_loss=0.0533, over 1422067.18 frames. ], batch size: 50, lr: 1.36e-02, grad_scale: 16.0 +2023-03-20 22:00:39,647 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31915.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:00:43,493 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 22:00:43,502 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 22:00:48,942 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 22:00:52,263 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31939.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:00:53,212 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 22:00:57,731 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31950.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:01:05,067 INFO [train.py:901] (1/2) Epoch 12, batch 900, loss[loss=0.1707, simple_loss=0.2446, pruned_loss=0.04844, over 7336.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2485, pruned_loss=0.05325, over 1425128.45 frames. ], batch size: 61, lr: 1.36e-02, grad_scale: 16.0 +2023-03-20 22:01:11,331 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31976.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:01:31,885 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9014, 4.0093, 3.7845, 4.0137, 3.9069, 3.9578, 4.1810, 4.3650], + device='cuda:1'), covar=tensor([0.0291, 0.0226, 0.0284, 0.0244, 0.0383, 0.0319, 0.0418, 0.0308], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0101, 0.0093, 0.0107, 0.0097, 0.0083, 0.0077, 0.0082], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:01:32,911 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32011.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:01:33,765 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.557e+02 2.399e+02 2.836e+02 3.405e+02 6.362e+02, threshold=5.672e+02, percent-clipped=3.0 +2023-03-20 22:01:34,345 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 22:01:34,801 INFO [train.py:901] (1/2) Epoch 12, batch 950, loss[loss=0.2054, simple_loss=0.2598, pruned_loss=0.0755, over 7214.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.248, pruned_loss=0.05326, over 1426864.86 frames. ], batch size: 45, lr: 1.36e-02, grad_scale: 16.0 +2023-03-20 22:01:46,685 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5637, 3.3010, 3.3274, 3.3802, 2.8156, 3.2265, 3.5660, 3.1438], + device='cuda:1'), covar=tensor([0.0228, 0.0220, 0.0270, 0.0254, 0.0766, 0.0224, 0.0367, 0.0282], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0063, 0.0066, 0.0052, 0.0101, 0.0069, 0.0068, 0.0065], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:01:57,302 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7800, 4.3756, 4.4552, 4.9774, 4.9578, 4.8927, 4.4373, 4.3258], + device='cuda:1'), covar=tensor([0.0823, 0.2265, 0.1949, 0.1102, 0.0646, 0.1342, 0.0610, 0.1166], + device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0286, 0.0231, 0.0224, 0.0171, 0.0289, 0.0155, 0.0202], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002], + device='cuda:1') +2023-03-20 22:01:58,838 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32060.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:01:59,265 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 22:02:01,268 INFO [train.py:901] (1/2) Epoch 12, batch 1000, loss[loss=0.1857, simple_loss=0.258, pruned_loss=0.05672, over 7364.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2479, pruned_loss=0.05323, over 1431701.61 frames. ], batch size: 73, lr: 1.36e-02, grad_scale: 16.0 +2023-03-20 22:02:01,363 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32065.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:02:19,917 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 22:02:25,231 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.567e+02 3.123e+02 3.891e+02 1.111e+03, threshold=6.246e+02, percent-clipped=6.0 +2023-03-20 22:02:25,321 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=32113.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:02:26,944 INFO [train.py:901] (1/2) Epoch 12, batch 1050, loss[loss=0.1843, simple_loss=0.259, pruned_loss=0.0548, over 7269.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2477, pruned_loss=0.05323, over 1433786.98 frames. ], batch size: 57, lr: 1.36e-02, grad_scale: 16.0 +2023-03-20 22:02:41,150 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 22:02:41,718 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32144.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:02:43,851 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32147.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 22:02:45,859 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2347, 1.4041, 1.5876, 1.1914, 2.0031, 1.4849, 1.6467, 1.3766], + device='cuda:1'), covar=tensor([0.0477, 0.0388, 0.0082, 0.0068, 0.0326, 0.0230, 0.0113, 0.0142], + device='cuda:1'), in_proj_covar=tensor([0.0022, 0.0020, 0.0021, 0.0019, 0.0019, 0.0019, 0.0020, 0.0020], + device='cuda:1'), out_proj_covar=tensor([5.3309e-05, 4.9316e-05, 4.7868e-05, 4.1578e-05, 4.8242e-05, 4.6563e-05, + 4.8702e-05, 5.1589e-05], device='cuda:1') +2023-03-20 22:02:46,705 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 22:02:52,641 INFO [train.py:901] (1/2) Epoch 12, batch 1100, loss[loss=0.1822, simple_loss=0.2482, pruned_loss=0.05813, over 7278.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2478, pruned_loss=0.05306, over 1436401.92 frames. ], batch size: 70, lr: 1.36e-02, grad_scale: 16.0 +2023-03-20 22:02:57,851 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9611, 2.3781, 2.2931, 3.2254, 1.3795, 3.3965, 1.1880, 2.9773], + device='cuda:1'), covar=tensor([0.0067, 0.0764, 0.1473, 0.0043, 0.4074, 0.0074, 0.1114, 0.0201], + device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0265, 0.0314, 0.0148, 0.0298, 0.0151, 0.0272, 0.0202], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 22:02:59,736 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32179.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:03:05,727 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5623, 3.6224, 3.5113, 3.5744, 3.4484, 3.6465, 3.8564, 3.9539], + device='cuda:1'), covar=tensor([0.0211, 0.0172, 0.0212, 0.0209, 0.0335, 0.0264, 0.0259, 0.0171], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0100, 0.0092, 0.0106, 0.0097, 0.0084, 0.0077, 0.0081], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:03:07,707 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=32195.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 22:03:14,768 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 22:03:15,279 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 22:03:17,754 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.242e+02 2.391e+02 2.835e+02 3.446e+02 6.409e+02, threshold=5.669e+02, percent-clipped=2.0 +2023-03-20 22:03:18,793 INFO [train.py:901] (1/2) Epoch 12, batch 1150, loss[loss=0.1798, simple_loss=0.2612, pruned_loss=0.04924, over 7218.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2487, pruned_loss=0.0535, over 1438972.24 frames. ], batch size: 93, lr: 1.36e-02, grad_scale: 16.0 +2023-03-20 22:03:28,480 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 22:03:29,017 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 22:03:29,076 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32234.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:03:33,464 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-03-20 22:03:44,687 INFO [train.py:901] (1/2) Epoch 12, batch 1200, loss[loss=0.161, simple_loss=0.238, pruned_loss=0.04199, over 7360.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2484, pruned_loss=0.05312, over 1438514.04 frames. ], batch size: 51, lr: 1.35e-02, grad_scale: 16.0 +2023-03-20 22:03:47,773 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32271.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:04:00,831 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 22:04:05,844 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32306.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:04:09,313 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.225e+02 2.639e+02 3.097e+02 5.486e+02, threshold=5.278e+02, percent-clipped=0.0 +2023-03-20 22:04:10,303 INFO [train.py:901] (1/2) Epoch 12, batch 1250, loss[loss=0.1696, simple_loss=0.2408, pruned_loss=0.04924, over 7162.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2477, pruned_loss=0.05283, over 1440213.77 frames. ], batch size: 41, lr: 1.35e-02, grad_scale: 16.0 +2023-03-20 22:04:25,369 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 22:04:29,294 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 22:04:30,795 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 22:04:33,465 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32360.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:04:35,932 INFO [train.py:901] (1/2) Epoch 12, batch 1300, loss[loss=0.1994, simple_loss=0.2649, pruned_loss=0.06693, over 7311.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2472, pruned_loss=0.05286, over 1439530.05 frames. ], batch size: 49, lr: 1.35e-02, grad_scale: 16.0 +2023-03-20 22:04:53,791 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 22:04:56,152 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 22:04:59,352 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=32408.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:04:59,810 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 22:05:01,805 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.570e+02 3.128e+02 3.930e+02 7.869e+02, threshold=6.255e+02, percent-clipped=7.0 +2023-03-20 22:05:02,826 INFO [train.py:901] (1/2) Epoch 12, batch 1350, loss[loss=0.164, simple_loss=0.2451, pruned_loss=0.04142, over 7332.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2479, pruned_loss=0.05293, over 1442184.54 frames. ], batch size: 44, lr: 1.35e-02, grad_scale: 16.0 +2023-03-20 22:05:10,309 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 22:05:16,434 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2573, 2.8611, 3.0082, 2.8578, 2.4336, 2.2802, 3.1178, 2.2085], + device='cuda:1'), covar=tensor([0.0233, 0.0220, 0.0214, 0.0222, 0.0281, 0.0413, 0.0252, 0.0765], + device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0274, 0.0233, 0.0286, 0.0291, 0.0292, 0.0267, 0.0282], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 22:05:17,055 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=6.05 vs. limit=5.0 +2023-03-20 22:05:17,402 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32444.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:05:17,738 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 +2023-03-20 22:05:21,053 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6707, 3.1066, 3.3599, 3.6667, 3.6277, 3.7745, 3.3649, 3.3843], + device='cuda:1'), covar=tensor([0.0035, 0.0118, 0.0058, 0.0045, 0.0049, 0.0035, 0.0071, 0.0068], + device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0042, 0.0039, 0.0037, 0.0037, 0.0038, 0.0044, 0.0045], + device='cuda:1'), out_proj_covar=tensor([7.9522e-05, 1.1951e-04, 1.0805e-04, 9.3935e-05, 9.5323e-05, 9.9022e-05, + 1.2329e-04, 1.1943e-04], device='cuda:1') +2023-03-20 22:05:28,602 INFO [train.py:901] (1/2) Epoch 12, batch 1400, loss[loss=0.1884, simple_loss=0.261, pruned_loss=0.05787, over 7222.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.248, pruned_loss=0.05247, over 1443501.06 frames. ], batch size: 93, lr: 1.35e-02, grad_scale: 16.0 +2023-03-20 22:05:35,680 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32479.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:05:42,673 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 22:05:42,731 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=32492.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:05:53,135 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.809e+02 2.399e+02 2.899e+02 3.458e+02 5.987e+02, threshold=5.799e+02, percent-clipped=0.0 +2023-03-20 22:05:54,229 INFO [train.py:901] (1/2) Epoch 12, batch 1450, loss[loss=0.1322, simple_loss=0.203, pruned_loss=0.03074, over 6942.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2482, pruned_loss=0.05228, over 1444493.75 frames. ], batch size: 35, lr: 1.35e-02, grad_scale: 16.0 +2023-03-20 22:06:00,332 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=32527.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:06:03,956 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32534.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:06:06,517 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 22:06:10,369 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 +2023-03-20 22:06:20,131 INFO [train.py:901] (1/2) Epoch 12, batch 1500, loss[loss=0.1778, simple_loss=0.2523, pruned_loss=0.05163, over 7286.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.248, pruned_loss=0.05239, over 1444835.13 frames. ], batch size: 77, lr: 1.35e-02, grad_scale: 16.0 +2023-03-20 22:06:22,745 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 22:06:23,755 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32571.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:06:29,201 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=32582.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:06:41,431 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32606.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:06:44,761 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.462e+02 2.853e+02 3.546e+02 6.447e+02, threshold=5.706e+02, percent-clipped=3.0 +2023-03-20 22:06:45,773 INFO [train.py:901] (1/2) Epoch 12, batch 1550, loss[loss=0.1594, simple_loss=0.2213, pruned_loss=0.04872, over 7173.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2494, pruned_loss=0.05307, over 1445571.70 frames. ], batch size: 39, lr: 1.35e-02, grad_scale: 16.0 +2023-03-20 22:06:46,771 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 22:06:47,779 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=32619.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:06:53,228 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.02 vs. limit=5.0 +2023-03-20 22:06:59,967 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 +2023-03-20 22:07:06,084 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=32654.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:07:12,100 INFO [train.py:901] (1/2) Epoch 12, batch 1600, loss[loss=0.175, simple_loss=0.2494, pruned_loss=0.05025, over 7335.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2485, pruned_loss=0.05309, over 1440853.98 frames. ], batch size: 61, lr: 1.35e-02, grad_scale: 16.0 +2023-03-20 22:07:18,125 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 22:07:19,111 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 22:07:22,220 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 22:07:24,083 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-20 22:07:31,718 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 22:07:36,878 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.457e+02 2.362e+02 2.862e+02 3.457e+02 8.164e+02, threshold=5.723e+02, percent-clipped=2.0 +2023-03-20 22:07:36,916 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 22:07:37,914 INFO [train.py:901] (1/2) Epoch 12, batch 1650, loss[loss=0.1942, simple_loss=0.2634, pruned_loss=0.06249, over 7301.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2475, pruned_loss=0.05257, over 1438973.71 frames. ], batch size: 86, lr: 1.34e-02, grad_scale: 16.0 +2023-03-20 22:07:42,381 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-03-20 22:07:44,889 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 22:07:47,799 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-20 22:07:49,677 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 +2023-03-20 22:07:58,156 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2004, 1.2363, 1.1058, 1.2494, 1.0359, 0.8785, 1.1469, 0.8712], + device='cuda:1'), covar=tensor([0.0099, 0.0103, 0.0173, 0.0091, 0.0076, 0.0086, 0.0096, 0.0131], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0019, 0.0020, 0.0018, 0.0020, 0.0018, 0.0019, 0.0025], + device='cuda:1'), out_proj_covar=tensor([2.4663e-05, 2.1856e-05, 2.4205e-05, 2.0123e-05, 2.5090e-05, 1.9951e-05, + 2.2876e-05, 3.1457e-05], device='cuda:1') +2023-03-20 22:08:02,094 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 22:08:03,576 INFO [train.py:901] (1/2) Epoch 12, batch 1700, loss[loss=0.1905, simple_loss=0.2596, pruned_loss=0.06066, over 7309.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2476, pruned_loss=0.05254, over 1441018.65 frames. ], batch size: 68, lr: 1.34e-02, grad_scale: 16.0 +2023-03-20 22:08:04,485 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-03-20 22:08:05,579 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 22:08:13,877 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-03-20 22:08:17,088 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 22:08:28,543 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.594e+02 2.417e+02 2.972e+02 3.822e+02 7.800e+02, threshold=5.944e+02, percent-clipped=3.0 +2023-03-20 22:08:29,528 INFO [train.py:901] (1/2) Epoch 12, batch 1750, loss[loss=0.1797, simple_loss=0.2482, pruned_loss=0.05561, over 7296.00 frames. ], tot_loss[loss=0.176, simple_loss=0.247, pruned_loss=0.05252, over 1441160.52 frames. ], batch size: 80, lr: 1.34e-02, grad_scale: 16.0 +2023-03-20 22:08:32,367 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 +2023-03-20 22:08:43,145 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 22:08:44,146 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 22:08:55,110 INFO [train.py:901] (1/2) Epoch 12, batch 1800, loss[loss=0.1681, simple_loss=0.2434, pruned_loss=0.0464, over 7317.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2469, pruned_loss=0.05231, over 1443045.53 frames. ], batch size: 80, lr: 1.34e-02, grad_scale: 16.0 +2023-03-20 22:09:05,604 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 22:09:14,430 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1899, 2.7763, 2.0813, 3.0420, 1.3623, 3.1496, 1.3536, 2.7261], + device='cuda:1'), covar=tensor([0.0059, 0.0786, 0.1892, 0.0048, 0.4732, 0.0058, 0.1089, 0.0124], + device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0264, 0.0305, 0.0147, 0.0292, 0.0149, 0.0267, 0.0204], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 22:09:14,887 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4815, 3.3739, 3.5231, 3.4414, 3.2857, 3.3259, 2.5121, 3.3222], + device='cuda:1'), covar=tensor([0.1029, 0.0353, 0.0707, 0.1662, 0.1635, 0.0787, 0.3033, 0.1543], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0035, 0.0034, 0.0036, 0.0032, 0.0030, 0.0046, 0.0034], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:09:15,463 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0323, 2.8826, 2.9438, 2.9247, 2.2756, 2.4673, 3.1846, 2.3914], + device='cuda:1'), covar=tensor([0.0203, 0.0201, 0.0250, 0.0250, 0.0334, 0.0455, 0.0259, 0.0723], + device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0283, 0.0242, 0.0295, 0.0297, 0.0296, 0.0279, 0.0291], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 22:09:18,364 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0083, 3.5956, 4.0391, 3.9282, 3.8810, 4.1216, 3.9339, 3.6707], + device='cuda:1'), covar=tensor([0.0037, 0.0091, 0.0040, 0.0038, 0.0039, 0.0029, 0.0042, 0.0056], + device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0042, 0.0038, 0.0036, 0.0038, 0.0038, 0.0044, 0.0046], + device='cuda:1'), out_proj_covar=tensor([7.9298e-05, 1.1940e-04, 1.0327e-04, 9.2592e-05, 9.5607e-05, 9.8824e-05, + 1.2172e-04, 1.1993e-04], device='cuda:1') +2023-03-20 22:09:19,255 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 22:09:19,727 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.412e+02 2.518e+02 3.025e+02 3.704e+02 1.165e+03, threshold=6.050e+02, percent-clipped=2.0 +2023-03-20 22:09:20,758 INFO [train.py:901] (1/2) Epoch 12, batch 1850, loss[loss=0.1772, simple_loss=0.2517, pruned_loss=0.05141, over 7345.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2463, pruned_loss=0.05204, over 1445290.44 frames. ], batch size: 54, lr: 1.34e-02, grad_scale: 16.0 +2023-03-20 22:09:29,921 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 22:09:46,114 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 22:09:46,601 INFO [train.py:901] (1/2) Epoch 12, batch 1900, loss[loss=0.1658, simple_loss=0.2383, pruned_loss=0.0467, over 7213.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2465, pruned_loss=0.05203, over 1443220.18 frames. ], batch size: 50, lr: 1.34e-02, grad_scale: 16.0 +2023-03-20 22:09:47,235 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32966.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:09:53,880 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.37 vs. limit=5.0 +2023-03-20 22:09:58,644 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-20 22:10:10,644 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 22:10:12,137 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.404e+02 2.470e+02 2.955e+02 3.740e+02 5.660e+02, threshold=5.911e+02, percent-clipped=0.0 +2023-03-20 22:10:13,146 INFO [train.py:901] (1/2) Epoch 12, batch 1950, loss[loss=0.1814, simple_loss=0.2549, pruned_loss=0.05391, over 7282.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.247, pruned_loss=0.05182, over 1445413.66 frames. ], batch size: 66, lr: 1.34e-02, grad_scale: 16.0 +2023-03-20 22:10:19,215 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33027.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:10:20,396 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-20 22:10:21,565 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 22:10:26,050 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 22:10:26,551 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 22:10:30,866 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 +2023-03-20 22:10:31,801 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-20 22:10:38,027 INFO [train.py:901] (1/2) Epoch 12, batch 2000, loss[loss=0.1586, simple_loss=0.24, pruned_loss=0.0386, over 7303.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2476, pruned_loss=0.05239, over 1444303.26 frames. ], batch size: 59, lr: 1.34e-02, grad_scale: 16.0 +2023-03-20 22:10:40,708 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33069.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:10:43,065 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 22:10:54,574 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 22:11:02,287 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 22:11:03,255 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.429e+02 3.002e+02 3.468e+02 9.132e+02, threshold=6.004e+02, percent-clipped=3.0 +2023-03-20 22:11:04,286 INFO [train.py:901] (1/2) Epoch 12, batch 2050, loss[loss=0.1883, simple_loss=0.2563, pruned_loss=0.06013, over 7277.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2482, pruned_loss=0.05254, over 1444390.09 frames. ], batch size: 52, lr: 1.34e-02, grad_scale: 16.0 +2023-03-20 22:11:11,880 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33130.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:11:17,657 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 +2023-03-20 22:11:24,399 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33155.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:11:25,497 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6215, 3.4841, 3.1134, 3.3660, 2.5106, 2.5734, 3.7365, 2.5762], + device='cuda:1'), covar=tensor([0.0168, 0.0212, 0.0268, 0.0226, 0.0308, 0.0438, 0.0245, 0.0715], + device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0280, 0.0237, 0.0289, 0.0291, 0.0290, 0.0275, 0.0287], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 22:11:29,948 INFO [train.py:901] (1/2) Epoch 12, batch 2100, loss[loss=0.1636, simple_loss=0.2445, pruned_loss=0.04133, over 7281.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2487, pruned_loss=0.05267, over 1444795.92 frames. ], batch size: 77, lr: 1.34e-02, grad_scale: 16.0 +2023-03-20 22:11:36,001 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 22:11:39,106 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 22:11:46,259 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8546, 4.0666, 3.9929, 4.1089, 3.9695, 4.1992, 4.4767, 4.4970], + device='cuda:1'), covar=tensor([0.0193, 0.0134, 0.0182, 0.0137, 0.0205, 0.0170, 0.0192, 0.0153], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0101, 0.0095, 0.0106, 0.0098, 0.0083, 0.0080, 0.0083], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:11:52,528 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4652, 4.9660, 5.0259, 4.9923, 4.8103, 4.5589, 5.0691, 4.8236], + device='cuda:1'), covar=tensor([0.0358, 0.0327, 0.0312, 0.0340, 0.0245, 0.0285, 0.0262, 0.0437], + device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0180, 0.0132, 0.0136, 0.0112, 0.0168, 0.0143, 0.0115], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:11:54,916 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.733e+02 2.641e+02 3.038e+02 3.805e+02 9.449e+02, threshold=6.075e+02, percent-clipped=6.0 +2023-03-20 22:11:55,931 INFO [train.py:901] (1/2) Epoch 12, batch 2150, loss[loss=0.1778, simple_loss=0.2534, pruned_loss=0.05115, over 7313.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2492, pruned_loss=0.05284, over 1444902.85 frames. ], batch size: 75, lr: 1.33e-02, grad_scale: 16.0 +2023-03-20 22:11:56,618 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33216.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:12:01,143 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1006, 3.1235, 1.9259, 3.4097, 2.3797, 2.9883, 1.7030, 1.8719], + device='cuda:1'), covar=tensor([0.0248, 0.0451, 0.1726, 0.0304, 0.0331, 0.0338, 0.2347, 0.1223], + device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0212, 0.0297, 0.0209, 0.0231, 0.0218, 0.0263, 0.0282], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 22:12:18,613 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9077, 3.9762, 3.8996, 3.9642, 3.6480, 4.0393, 4.3136, 4.3734], + device='cuda:1'), covar=tensor([0.0184, 0.0153, 0.0153, 0.0172, 0.0323, 0.0179, 0.0185, 0.0137], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0102, 0.0096, 0.0107, 0.0100, 0.0084, 0.0080, 0.0083], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:12:21,547 INFO [train.py:901] (1/2) Epoch 12, batch 2200, loss[loss=0.1722, simple_loss=0.2472, pruned_loss=0.04859, over 7217.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.249, pruned_loss=0.05272, over 1445271.34 frames. ], batch size: 93, lr: 1.33e-02, grad_scale: 16.0 +2023-03-20 22:12:26,162 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 22:12:46,182 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.353e+02 2.489e+02 2.908e+02 3.489e+02 9.898e+02, threshold=5.816e+02, percent-clipped=3.0 +2023-03-20 22:12:47,226 INFO [train.py:901] (1/2) Epoch 12, batch 2250, loss[loss=0.1873, simple_loss=0.2587, pruned_loss=0.058, over 7295.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2479, pruned_loss=0.0522, over 1443410.76 frames. ], batch size: 86, lr: 1.33e-02, grad_scale: 16.0 +2023-03-20 22:12:50,790 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33322.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:12:59,924 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 22:13:00,435 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 22:13:12,059 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7483, 5.1757, 5.1953, 5.1583, 5.0036, 4.7445, 5.2949, 5.0184], + device='cuda:1'), covar=tensor([0.0332, 0.0335, 0.0357, 0.0387, 0.0257, 0.0262, 0.0249, 0.0415], + device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0180, 0.0130, 0.0133, 0.0111, 0.0168, 0.0141, 0.0114], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:13:12,101 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3901, 4.2679, 3.7803, 3.6028, 4.0572, 2.4333, 1.7825, 4.3563], + device='cuda:1'), covar=tensor([0.0016, 0.0048, 0.0066, 0.0060, 0.0042, 0.0368, 0.0518, 0.0032], + device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0061, 0.0081, 0.0070, 0.0086, 0.0106, 0.0107, 0.0073], + device='cuda:1'), out_proj_covar=tensor([8.5935e-05, 9.0594e-05, 1.0830e-04, 1.0007e-04, 1.1661e-04, 1.4474e-04, + 1.4774e-04, 9.9807e-05], device='cuda:1') +2023-03-20 22:13:13,450 INFO [train.py:901] (1/2) Epoch 12, batch 2300, loss[loss=0.1775, simple_loss=0.2574, pruned_loss=0.04885, over 7333.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2473, pruned_loss=0.05204, over 1440960.77 frames. ], batch size: 63, lr: 1.33e-02, grad_scale: 16.0 +2023-03-20 22:13:13,470 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 22:13:14,380 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 +2023-03-20 22:13:15,754 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.00 vs. limit=5.0 +2023-03-20 22:13:20,543 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.8251, 4.6190, 4.5549, 5.0093, 5.0046, 5.0153, 4.2131, 4.5680], + device='cuda:1'), covar=tensor([0.0856, 0.2244, 0.1875, 0.0898, 0.0632, 0.1189, 0.0673, 0.1002], + device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0284, 0.0231, 0.0223, 0.0165, 0.0293, 0.0159, 0.0201], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:13:37,841 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.704e+02 2.453e+02 2.925e+02 3.664e+02 6.582e+02, threshold=5.850e+02, percent-clipped=1.0 +2023-03-20 22:13:37,978 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.9395, 4.3457, 4.7007, 4.5143, 4.4543, 4.5643, 4.7619, 4.4980], + device='cuda:1'), covar=tensor([0.0058, 0.0078, 0.0063, 0.0067, 0.0198, 0.0056, 0.0097, 0.0059], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0065, 0.0067, 0.0054, 0.0108, 0.0071, 0.0069, 0.0067], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:13:38,898 INFO [train.py:901] (1/2) Epoch 12, batch 2350, loss[loss=0.1869, simple_loss=0.2575, pruned_loss=0.05813, over 7286.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2479, pruned_loss=0.05232, over 1440257.85 frames. ], batch size: 68, lr: 1.33e-02, grad_scale: 16.0 +2023-03-20 22:13:44,675 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33425.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:13:45,646 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5059, 3.9489, 4.0335, 3.9667, 3.8561, 4.0938, 4.1776, 3.8105], + device='cuda:1'), covar=tensor([0.0075, 0.0115, 0.0126, 0.0106, 0.0311, 0.0082, 0.0128, 0.0143], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0065, 0.0067, 0.0053, 0.0108, 0.0070, 0.0068, 0.0067], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:13:59,834 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 22:14:05,406 INFO [train.py:901] (1/2) Epoch 12, batch 2400, loss[loss=0.1696, simple_loss=0.2444, pruned_loss=0.04737, over 7253.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2476, pruned_loss=0.05193, over 1439432.30 frames. ], batch size: 55, lr: 1.33e-02, grad_scale: 16.0 +2023-03-20 22:14:05,427 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 22:14:09,008 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33472.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:14:15,938 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 22:14:18,462 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 22:14:24,021 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1587, 3.8808, 3.9584, 4.3979, 4.4205, 4.3451, 3.9233, 3.7647], + device='cuda:1'), covar=tensor([0.0893, 0.2388, 0.1870, 0.0958, 0.0633, 0.1275, 0.0742, 0.1200], + device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0289, 0.0237, 0.0226, 0.0169, 0.0294, 0.0159, 0.0204], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:14:29,178 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33511.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:14:30,081 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.498e+02 2.266e+02 2.777e+02 3.531e+02 6.731e+02, threshold=5.555e+02, percent-clipped=2.0 +2023-03-20 22:14:31,088 INFO [train.py:901] (1/2) Epoch 12, batch 2450, loss[loss=0.2125, simple_loss=0.2839, pruned_loss=0.07055, over 6688.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2482, pruned_loss=0.05239, over 1439909.36 frames. ], batch size: 106, lr: 1.33e-02, grad_scale: 16.0 +2023-03-20 22:14:40,721 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33533.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:14:44,584 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 22:14:52,289 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7707, 4.0361, 3.8529, 4.0082, 3.7079, 4.1792, 4.3217, 4.4617], + device='cuda:1'), covar=tensor([0.0217, 0.0143, 0.0176, 0.0151, 0.0273, 0.0135, 0.0277, 0.0154], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0099, 0.0094, 0.0105, 0.0097, 0.0081, 0.0077, 0.0080], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:14:56,694 INFO [train.py:901] (1/2) Epoch 12, batch 2500, loss[loss=0.1687, simple_loss=0.2451, pruned_loss=0.04615, over 7266.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2479, pruned_loss=0.0525, over 1439571.08 frames. ], batch size: 64, lr: 1.33e-02, grad_scale: 32.0 +2023-03-20 22:14:57,306 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5608, 3.9437, 4.0284, 3.9627, 3.7915, 3.8332, 3.9741, 3.8469], + device='cuda:1'), covar=tensor([0.0073, 0.0118, 0.0116, 0.0116, 0.0307, 0.0104, 0.0167, 0.0112], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0064, 0.0065, 0.0054, 0.0107, 0.0070, 0.0068, 0.0066], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:14:58,975 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-20 22:15:10,893 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 22:15:12,623 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 +2023-03-20 22:15:22,702 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.564e+02 2.335e+02 2.794e+02 3.571e+02 7.375e+02, threshold=5.588e+02, percent-clipped=5.0 +2023-03-20 22:15:23,265 INFO [train.py:901] (1/2) Epoch 12, batch 2550, loss[loss=0.1445, simple_loss=0.2034, pruned_loss=0.04275, over 6259.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2474, pruned_loss=0.05214, over 1440814.61 frames. ], batch size: 27, lr: 1.33e-02, grad_scale: 16.0 +2023-03-20 22:15:26,845 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33622.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:15:32,128 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 +2023-03-20 22:15:48,231 INFO [train.py:901] (1/2) Epoch 12, batch 2600, loss[loss=0.202, simple_loss=0.2609, pruned_loss=0.07155, over 7278.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2476, pruned_loss=0.05237, over 1440868.92 frames. ], batch size: 70, lr: 1.33e-02, grad_scale: 16.0 +2023-03-20 22:15:50,774 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=33670.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:15:54,328 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33677.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:16:12,797 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 2.314e+02 2.858e+02 3.397e+02 6.641e+02, threshold=5.717e+02, percent-clipped=1.0 +2023-03-20 22:16:13,301 INFO [train.py:901] (1/2) Epoch 12, batch 2650, loss[loss=0.1587, simple_loss=0.2221, pruned_loss=0.0477, over 7263.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2476, pruned_loss=0.0526, over 1441904.02 frames. ], batch size: 64, lr: 1.33e-02, grad_scale: 16.0 +2023-03-20 22:16:18,272 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33725.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:16:24,687 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33738.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 22:16:37,749 INFO [train.py:901] (1/2) Epoch 12, batch 2700, loss[loss=0.1474, simple_loss=0.2101, pruned_loss=0.04234, over 6993.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2474, pruned_loss=0.05267, over 1442331.61 frames. ], batch size: 35, lr: 1.32e-02, grad_scale: 16.0 +2023-03-20 22:16:41,731 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=33773.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:17:01,375 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33811.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:17:02,643 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.411e+02 2.993e+02 3.645e+02 8.242e+02, threshold=5.986e+02, percent-clipped=5.0 +2023-03-20 22:17:03,128 INFO [train.py:901] (1/2) Epoch 12, batch 2750, loss[loss=0.1805, simple_loss=0.2553, pruned_loss=0.05286, over 7324.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.247, pruned_loss=0.05258, over 1442940.14 frames. ], batch size: 83, lr: 1.32e-02, grad_scale: 16.0 +2023-03-20 22:17:09,562 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33828.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:17:12,100 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8179, 3.5326, 3.4794, 3.6222, 2.7849, 2.8803, 3.7765, 2.9663], + device='cuda:1'), covar=tensor([0.0157, 0.0171, 0.0188, 0.0187, 0.0336, 0.0410, 0.0194, 0.0622], + device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0278, 0.0239, 0.0289, 0.0292, 0.0291, 0.0279, 0.0286], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 22:17:14,010 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5415, 3.3810, 3.1486, 3.4562, 2.5123, 2.5165, 3.6905, 2.6165], + device='cuda:1'), covar=tensor([0.0192, 0.0227, 0.0303, 0.0238, 0.0405, 0.0575, 0.0228, 0.1009], + device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0278, 0.0238, 0.0289, 0.0291, 0.0290, 0.0278, 0.0286], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-20 22:17:19,334 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9994, 3.6219, 3.6304, 3.6039, 3.5666, 3.6346, 3.7750, 3.5487], + device='cuda:1'), covar=tensor([0.0098, 0.0150, 0.0131, 0.0160, 0.0326, 0.0096, 0.0161, 0.0137], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0065, 0.0066, 0.0055, 0.0108, 0.0071, 0.0069, 0.0067], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:17:24,658 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=33859.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:17:25,193 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8212, 3.5639, 3.4153, 3.5665, 2.9513, 3.5188, 3.5168, 3.4479], + device='cuda:1'), covar=tensor([0.0258, 0.0202, 0.0249, 0.0266, 0.0790, 0.0179, 0.0379, 0.0251], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0065, 0.0066, 0.0055, 0.0107, 0.0070, 0.0068, 0.0067], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:17:27,448 INFO [train.py:901] (1/2) Epoch 12, batch 2800, loss[loss=0.1658, simple_loss=0.2408, pruned_loss=0.04544, over 7265.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2472, pruned_loss=0.05277, over 1444280.26 frames. ], batch size: 57, lr: 1.32e-02, grad_scale: 16.0 +2023-03-20 22:17:30,612 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.00 vs. limit=2.0 +2023-03-20 22:17:52,812 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 22:17:52,865 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 22:17:52,889 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 22:17:52,906 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 22:17:53,078 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 22:17:53,092 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 22:17:53,104 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 22:17:53,199 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 22:17:53,273 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 22:17:53,287 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 22:17:53,303 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 22:17:53,777 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 22:17:53,783 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 22:17:53,791 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 22:17:53,926 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 22:17:53,929 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 22:17:53,986 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 22:18:01,878 INFO [train.py:901] (1/2) Epoch 13, batch 0, loss[loss=0.1845, simple_loss=0.2541, pruned_loss=0.05747, over 7315.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2541, pruned_loss=0.05747, over 7315.00 frames. ], batch size: 75, lr: 1.28e-02, grad_scale: 16.0 +2023-03-20 22:18:01,878 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 22:18:16,812 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1565, 4.1898, 4.1400, 4.7058, 4.6551, 4.6230, 4.2323, 4.1014], + device='cuda:1'), covar=tensor([0.0979, 0.1915, 0.2036, 0.0933, 0.0528, 0.1181, 0.0540, 0.0944], + device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0283, 0.0231, 0.0220, 0.0164, 0.0286, 0.0159, 0.0199], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:18:25,451 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2158, 2.4010, 2.0628, 3.0751, 1.7209, 2.4245, 2.2956, 1.6633], + device='cuda:1'), covar=tensor([0.1400, 0.0782, 0.2789, 0.0481, 0.0058, 0.0038, 0.0102, 0.0093], + device='cuda:1'), in_proj_covar=tensor([0.0286, 0.0227, 0.0272, 0.0248, 0.0127, 0.0119, 0.0142, 0.0158], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:18:27,578 INFO [train.py:935] (1/2) Epoch 13, validation: loss=0.1711, simple_loss=0.259, pruned_loss=0.04159, over 1622729.00 frames. +2023-03-20 22:18:27,578 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-20 22:18:34,106 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 22:18:37,998 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.83 vs. limit=2.0 +2023-03-20 22:18:40,145 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.660e+02 2.486e+02 3.069e+02 3.797e+02 7.499e+02, threshold=6.139e+02, percent-clipped=2.0 +2023-03-20 22:18:44,658 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 22:18:51,657 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-03-20 22:18:52,393 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 22:18:53,325 INFO [train.py:901] (1/2) Epoch 13, batch 50, loss[loss=0.1584, simple_loss=0.2365, pruned_loss=0.04014, over 7281.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2472, pruned_loss=0.05053, over 326509.04 frames. ], batch size: 70, lr: 1.28e-02, grad_scale: 16.0 +2023-03-20 22:18:53,486 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33939.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:18:54,361 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 22:18:56,909 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 22:19:01,996 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2388, 2.6389, 2.1595, 2.8049, 2.4951, 2.0461, 2.6778, 2.6406], + device='cuda:1'), covar=tensor([0.1118, 0.1070, 0.1216, 0.0858, 0.1381, 0.1045, 0.0545, 0.0802], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0042, 0.0046, 0.0039, 0.0042, 0.0040, 0.0043, 0.0037], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 22:19:11,097 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3211, 1.3520, 1.3059, 1.3909, 1.1425, 1.0061, 1.0667, 0.8455], + device='cuda:1'), covar=tensor([0.0102, 0.0138, 0.0129, 0.0047, 0.0083, 0.0102, 0.0086, 0.0107], + device='cuda:1'), in_proj_covar=tensor([0.0022, 0.0020, 0.0019, 0.0019, 0.0021, 0.0019, 0.0020, 0.0026], + device='cuda:1'), out_proj_covar=tensor([2.6103e-05, 2.2933e-05, 2.3833e-05, 2.0890e-05, 2.6005e-05, 2.1094e-05, + 2.3418e-05, 3.1415e-05], device='cuda:1') +2023-03-20 22:19:13,390 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 22:19:13,907 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 22:19:18,814 INFO [train.py:901] (1/2) Epoch 13, batch 100, loss[loss=0.168, simple_loss=0.2446, pruned_loss=0.04565, over 7296.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2474, pruned_loss=0.05124, over 574383.73 frames. ], batch size: 68, lr: 1.28e-02, grad_scale: 16.0 +2023-03-20 22:19:20,963 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8002, 3.7365, 3.0446, 3.0938, 3.0783, 2.3978, 1.6523, 3.7807], + device='cuda:1'), covar=tensor([0.0023, 0.0029, 0.0074, 0.0053, 0.0082, 0.0324, 0.0496, 0.0034], + device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0063, 0.0081, 0.0069, 0.0084, 0.0105, 0.0108, 0.0075], + device='cuda:1'), out_proj_covar=tensor([8.8069e-05, 9.1591e-05, 1.0843e-04, 1.0017e-04, 1.1330e-04, 1.4352e-04, + 1.4771e-04, 1.0160e-04], device='cuda:1') +2023-03-20 22:19:24,773 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34000.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:19:26,396 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-20 22:19:29,986 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.01 vs. limit=2.0 +2023-03-20 22:19:32,285 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.611e+02 2.332e+02 2.851e+02 3.491e+02 5.832e+02, threshold=5.703e+02, percent-clipped=0.0 +2023-03-20 22:19:41,879 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34033.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 22:19:44,839 INFO [train.py:901] (1/2) Epoch 13, batch 150, loss[loss=0.1737, simple_loss=0.2461, pruned_loss=0.05067, over 7354.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2466, pruned_loss=0.05125, over 767230.96 frames. ], batch size: 54, lr: 1.27e-02, grad_scale: 16.0 +2023-03-20 22:19:47,496 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34044.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:20:04,718 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5794, 3.8790, 4.1298, 4.1605, 3.8516, 4.0707, 4.2444, 3.8322], + device='cuda:1'), covar=tensor([0.0086, 0.0122, 0.0131, 0.0091, 0.0344, 0.0097, 0.0161, 0.0130], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0065, 0.0065, 0.0053, 0.0106, 0.0071, 0.0069, 0.0067], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:20:09,310 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7528, 3.4599, 3.5864, 3.8026, 3.5866, 3.7540, 3.7125, 3.6086], + device='cuda:1'), covar=tensor([0.0027, 0.0075, 0.0039, 0.0035, 0.0035, 0.0032, 0.0039, 0.0044], + device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0043, 0.0039, 0.0037, 0.0038, 0.0039, 0.0044, 0.0047], + device='cuda:1'), out_proj_covar=tensor([7.9262e-05, 1.1994e-04, 1.0926e-04, 9.2318e-05, 9.5452e-05, 9.8182e-05, + 1.2027e-04, 1.2244e-04], device='cuda:1') +2023-03-20 22:20:10,703 INFO [train.py:901] (1/2) Epoch 13, batch 200, loss[loss=0.1662, simple_loss=0.2427, pruned_loss=0.04482, over 7314.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.247, pruned_loss=0.05113, over 918692.24 frames. ], batch size: 83, lr: 1.27e-02, grad_scale: 16.0 +2023-03-20 22:20:14,693 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 22:20:19,306 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34105.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:20:20,169 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 22:20:22,278 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1001, 3.5792, 3.7162, 3.7154, 3.5038, 3.6211, 3.8230, 3.5858], + device='cuda:1'), covar=tensor([0.0094, 0.0148, 0.0107, 0.0113, 0.0318, 0.0106, 0.0176, 0.0102], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0065, 0.0065, 0.0053, 0.0106, 0.0070, 0.0068, 0.0067], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:20:23,658 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.373e+02 2.823e+02 3.333e+02 8.127e+02, threshold=5.645e+02, percent-clipped=4.0 +2023-03-20 22:20:26,731 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 22:20:30,819 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34128.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:20:36,198 INFO [train.py:901] (1/2) Epoch 13, batch 250, loss[loss=0.18, simple_loss=0.2483, pruned_loss=0.05588, over 7304.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2462, pruned_loss=0.05093, over 1033615.34 frames. ], batch size: 80, lr: 1.27e-02, grad_scale: 16.0 +2023-03-20 22:20:39,271 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 22:20:39,889 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34145.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:20:55,421 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=34176.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:20:59,407 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 22:21:02,454 INFO [train.py:901] (1/2) Epoch 13, batch 300, loss[loss=0.2022, simple_loss=0.2733, pruned_loss=0.06551, over 6775.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2451, pruned_loss=0.05077, over 1122324.55 frames. ], batch size: 107, lr: 1.27e-02, grad_scale: 16.0 +2023-03-20 22:21:07,587 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9368, 3.7615, 3.8136, 3.7218, 3.7909, 3.8125, 3.9017, 3.7216], + device='cuda:1'), covar=tensor([0.0027, 0.0059, 0.0030, 0.0056, 0.0029, 0.0028, 0.0033, 0.0048], + device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0043, 0.0039, 0.0037, 0.0038, 0.0039, 0.0044, 0.0046], + device='cuda:1'), out_proj_covar=tensor([7.7417e-05, 1.1966e-04, 1.0702e-04, 9.2540e-05, 9.5285e-05, 9.8496e-05, + 1.2041e-04, 1.2124e-04], device='cuda:1') +2023-03-20 22:21:09,344 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 22:21:11,443 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34206.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:21:15,223 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.521e+02 2.911e+02 3.518e+02 6.846e+02, threshold=5.822e+02, percent-clipped=2.0 +2023-03-20 22:21:22,309 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5343, 2.7433, 2.5273, 3.5115, 1.6449, 3.4773, 1.4006, 3.0528], + device='cuda:1'), covar=tensor([0.0071, 0.0590, 0.1350, 0.0041, 0.3828, 0.0061, 0.0977, 0.0158], + device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0257, 0.0296, 0.0144, 0.0283, 0.0149, 0.0262, 0.0202], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 22:21:28,194 INFO [train.py:901] (1/2) Epoch 13, batch 350, loss[loss=0.169, simple_loss=0.2507, pruned_loss=0.04368, over 7280.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2449, pruned_loss=0.05067, over 1195777.97 frames. ], batch size: 77, lr: 1.27e-02, grad_scale: 16.0 +2023-03-20 22:21:43,293 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 22:21:43,962 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6875, 1.9731, 2.0363, 2.9432, 1.4256, 2.5970, 1.1417, 2.6812], + device='cuda:1'), covar=tensor([0.0105, 0.0885, 0.1731, 0.0054, 0.4098, 0.0071, 0.1023, 0.0176], + device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0259, 0.0300, 0.0146, 0.0286, 0.0150, 0.0264, 0.0206], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 22:21:53,954 INFO [train.py:901] (1/2) Epoch 13, batch 400, loss[loss=0.1853, simple_loss=0.248, pruned_loss=0.06133, over 7350.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2454, pruned_loss=0.05094, over 1252233.66 frames. ], batch size: 54, lr: 1.27e-02, grad_scale: 16.0 +2023-03-20 22:21:57,094 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34295.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:22:06,468 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.410e+02 2.808e+02 3.305e+02 7.616e+02, threshold=5.616e+02, percent-clipped=2.0 +2023-03-20 22:22:16,638 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34333.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:22:18,439 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-20 22:22:19,532 INFO [train.py:901] (1/2) Epoch 13, batch 450, loss[loss=0.1677, simple_loss=0.2421, pruned_loss=0.04663, over 7227.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.245, pruned_loss=0.05035, over 1293966.96 frames. ], batch size: 50, lr: 1.27e-02, grad_scale: 16.0 +2023-03-20 22:22:24,684 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 22:22:25,199 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 22:22:38,072 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1751, 3.6485, 3.8131, 3.7316, 3.6154, 3.6714, 3.9224, 3.5870], + device='cuda:1'), covar=tensor([0.0084, 0.0146, 0.0092, 0.0141, 0.0298, 0.0100, 0.0141, 0.0111], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0065, 0.0065, 0.0055, 0.0107, 0.0071, 0.0068, 0.0067], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:22:41,065 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.1421, 5.6895, 5.7274, 5.6451, 5.4292, 5.3394, 5.7672, 5.5689], + device='cuda:1'), covar=tensor([0.0335, 0.0269, 0.0332, 0.0355, 0.0271, 0.0219, 0.0271, 0.0399], + device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0179, 0.0132, 0.0132, 0.0111, 0.0166, 0.0143, 0.0114], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:22:41,571 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=34381.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:22:45,508 INFO [train.py:901] (1/2) Epoch 13, batch 500, loss[loss=0.1747, simple_loss=0.2507, pruned_loss=0.04935, over 7294.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2451, pruned_loss=0.05056, over 1326265.63 frames. ], batch size: 77, lr: 1.27e-02, grad_scale: 16.0 +2023-03-20 22:22:51,366 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34400.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:22:58,935 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.527e+02 2.514e+02 3.021e+02 3.896e+02 7.702e+02, threshold=6.043e+02, percent-clipped=7.0 +2023-03-20 22:22:58,981 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 22:23:01,003 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 22:23:01,503 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 22:23:03,957 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 22:23:08,012 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 22:23:11,458 INFO [train.py:901] (1/2) Epoch 13, batch 550, loss[loss=0.1597, simple_loss=0.2348, pruned_loss=0.04229, over 7367.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2455, pruned_loss=0.05078, over 1353818.55 frames. ], batch size: 44, lr: 1.27e-02, grad_scale: 16.0 +2023-03-20 22:23:19,202 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 22:23:27,108 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 22:23:27,705 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6283, 2.4106, 2.9265, 2.7171, 3.0021, 2.8089, 2.3873, 2.8205], + device='cuda:1'), covar=tensor([0.1773, 0.0336, 0.1399, 0.1604, 0.0821, 0.0854, 0.2308, 0.1523], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0037, 0.0035, 0.0036, 0.0032, 0.0032, 0.0047, 0.0035], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:23:30,149 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 22:23:37,757 INFO [train.py:901] (1/2) Epoch 13, batch 600, loss[loss=0.1727, simple_loss=0.2409, pruned_loss=0.05229, over 7268.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2453, pruned_loss=0.05078, over 1373460.13 frames. ], batch size: 52, lr: 1.27e-02, grad_scale: 16.0 +2023-03-20 22:23:38,265 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 22:23:43,836 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34501.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:23:50,211 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.492e+02 2.326e+02 2.723e+02 3.243e+02 6.434e+02, threshold=5.446e+02, percent-clipped=1.0 +2023-03-20 22:23:53,285 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 22:23:59,947 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3008, 3.1328, 3.0403, 3.2835, 2.5061, 2.4449, 3.3501, 2.4978], + device='cuda:1'), covar=tensor([0.0202, 0.0237, 0.0233, 0.0262, 0.0322, 0.0418, 0.0272, 0.0889], + device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0288, 0.0244, 0.0297, 0.0299, 0.0302, 0.0287, 0.0294], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 22:24:01,511 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1929, 1.6822, 1.9757, 1.8348, 1.5711, 1.4137, 1.6987, 1.3761], + device='cuda:1'), covar=tensor([0.0327, 0.0277, 0.0110, 0.0061, 0.0753, 0.0786, 0.0140, 0.0236], + device='cuda:1'), in_proj_covar=tensor([0.0021, 0.0021, 0.0020, 0.0019, 0.0021, 0.0021, 0.0020, 0.0022], + device='cuda:1'), out_proj_covar=tensor([5.3720e-05, 5.1215e-05, 4.7153e-05, 4.3565e-05, 5.2936e-05, 5.0032e-05, + 4.9741e-05, 5.5096e-05], device='cuda:1') +2023-03-20 22:24:01,836 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 22:24:03,306 INFO [train.py:901] (1/2) Epoch 13, batch 650, loss[loss=0.144, simple_loss=0.2169, pruned_loss=0.0355, over 7135.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2453, pruned_loss=0.05069, over 1388631.19 frames. ], batch size: 41, lr: 1.27e-02, grad_scale: 16.0 +2023-03-20 22:24:18,765 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 22:24:20,379 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2141, 1.4183, 1.3498, 1.4261, 1.3765, 1.1556, 1.3507, 0.9742], + device='cuda:1'), covar=tensor([0.0317, 0.0135, 0.0204, 0.0115, 0.0219, 0.0140, 0.0171, 0.0213], + device='cuda:1'), in_proj_covar=tensor([0.0022, 0.0020, 0.0020, 0.0019, 0.0022, 0.0020, 0.0020, 0.0026], + device='cuda:1'), out_proj_covar=tensor([2.6435e-05, 2.2900e-05, 2.4771e-05, 2.1285e-05, 2.6520e-05, 2.2413e-05, + 2.3822e-05, 3.1893e-05], device='cuda:1') +2023-03-20 22:24:28,281 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 22:24:28,756 INFO [train.py:901] (1/2) Epoch 13, batch 700, loss[loss=0.1592, simple_loss=0.2341, pruned_loss=0.04216, over 7284.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2448, pruned_loss=0.05052, over 1401617.00 frames. ], batch size: 70, lr: 1.26e-02, grad_scale: 16.0 +2023-03-20 22:24:30,914 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34593.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:24:31,917 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34595.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:24:41,583 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.640e+02 2.349e+02 2.874e+02 3.606e+02 6.816e+02, threshold=5.747e+02, percent-clipped=3.0 +2023-03-20 22:24:42,420 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-03-20 22:24:46,323 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34622.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:24:48,323 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34626.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:24:52,716 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 22:24:52,730 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 22:24:54,698 INFO [train.py:901] (1/2) Epoch 13, batch 750, loss[loss=0.176, simple_loss=0.2467, pruned_loss=0.05263, over 7299.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2445, pruned_loss=0.05031, over 1410551.61 frames. ], batch size: 86, lr: 1.26e-02, grad_scale: 16.0 +2023-03-20 22:24:56,757 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=34643.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:25:02,433 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34654.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:25:07,461 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 22:25:11,957 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 22:25:17,568 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34683.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 22:25:18,417 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 22:25:19,429 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 22:25:19,550 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34687.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:25:20,413 INFO [train.py:901] (1/2) Epoch 13, batch 800, loss[loss=0.1813, simple_loss=0.2453, pruned_loss=0.05862, over 7298.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2444, pruned_loss=0.05015, over 1415910.81 frames. ], batch size: 49, lr: 1.26e-02, grad_scale: 16.0 +2023-03-20 22:25:25,937 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4439, 3.6493, 3.4975, 3.5762, 3.3206, 3.6916, 3.9860, 4.0265], + device='cuda:1'), covar=tensor([0.0250, 0.0158, 0.0240, 0.0208, 0.0344, 0.0206, 0.0223, 0.0178], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0103, 0.0098, 0.0105, 0.0098, 0.0082, 0.0079, 0.0082], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:25:25,960 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34700.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:25:30,310 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 22:25:31,611 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34710.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 22:25:33,443 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.193e+02 2.697e+02 3.424e+02 6.128e+02, threshold=5.393e+02, percent-clipped=1.0 +2023-03-20 22:25:45,934 INFO [train.py:901] (1/2) Epoch 13, batch 850, loss[loss=0.1704, simple_loss=0.2499, pruned_loss=0.04545, over 7246.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2445, pruned_loss=0.05006, over 1422631.16 frames. ], batch size: 55, lr: 1.26e-02, grad_scale: 16.0 +2023-03-20 22:25:50,370 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 22:25:50,379 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 22:25:50,413 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=34748.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:25:50,482 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4165, 4.3270, 3.8768, 3.5502, 3.7973, 2.5204, 1.5560, 4.2493], + device='cuda:1'), covar=tensor([0.0020, 0.0024, 0.0051, 0.0065, 0.0055, 0.0373, 0.0542, 0.0042], + device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0064, 0.0084, 0.0073, 0.0090, 0.0110, 0.0114, 0.0079], + device='cuda:1'), out_proj_covar=tensor([9.1042e-05, 9.3762e-05, 1.1212e-04, 1.0483e-04, 1.2086e-04, 1.4958e-04, + 1.5516e-04, 1.0539e-04], device='cuda:1') +2023-03-20 22:25:56,390 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 22:25:59,906 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 22:26:02,534 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34771.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 22:26:07,183 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.73 vs. limit=5.0 +2023-03-20 22:26:11,304 INFO [train.py:901] (1/2) Epoch 13, batch 900, loss[loss=0.1669, simple_loss=0.2481, pruned_loss=0.04286, over 7281.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2447, pruned_loss=0.04988, over 1427228.41 frames. ], batch size: 57, lr: 1.26e-02, grad_scale: 16.0 +2023-03-20 22:26:17,776 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34801.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:26:24,749 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 2.135e+02 2.636e+02 3.552e+02 6.481e+02, threshold=5.271e+02, percent-clipped=4.0 +2023-03-20 22:26:34,238 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-20 22:26:37,305 INFO [train.py:901] (1/2) Epoch 13, batch 950, loss[loss=0.2145, simple_loss=0.2855, pruned_loss=0.07173, over 6712.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2448, pruned_loss=0.05019, over 1426441.67 frames. ], batch size: 106, lr: 1.26e-02, grad_scale: 16.0 +2023-03-20 22:26:39,368 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 22:26:42,916 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=34849.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:26:55,054 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0353, 2.3355, 2.0386, 3.2916, 1.5313, 3.4965, 1.4367, 2.8663], + device='cuda:1'), covar=tensor([0.0051, 0.0776, 0.1521, 0.0046, 0.3501, 0.0063, 0.0900, 0.0148], + device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0260, 0.0296, 0.0147, 0.0286, 0.0150, 0.0265, 0.0204], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 22:27:01,849 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 22:27:02,796 INFO [train.py:901] (1/2) Epoch 13, batch 1000, loss[loss=0.1907, simple_loss=0.2589, pruned_loss=0.06128, over 7260.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2454, pruned_loss=0.05059, over 1430303.04 frames. ], batch size: 89, lr: 1.26e-02, grad_scale: 16.0 +2023-03-20 22:27:15,916 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.342e+02 2.774e+02 3.398e+02 6.564e+02, threshold=5.548e+02, percent-clipped=3.0 +2023-03-20 22:27:22,402 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 22:27:24,203 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.11 vs. limit=2.0 +2023-03-20 22:27:28,811 INFO [train.py:901] (1/2) Epoch 13, batch 1050, loss[loss=0.183, simple_loss=0.2524, pruned_loss=0.05679, over 7280.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2454, pruned_loss=0.05072, over 1433845.85 frames. ], batch size: 77, lr: 1.26e-02, grad_scale: 16.0 +2023-03-20 22:27:33,963 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34949.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:27:43,252 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 22:27:44,047 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 +2023-03-20 22:27:48,348 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 22:27:48,892 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34978.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 22:27:50,904 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34982.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:27:54,363 INFO [train.py:901] (1/2) Epoch 13, batch 1100, loss[loss=0.1666, simple_loss=0.2437, pruned_loss=0.04471, over 7262.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2453, pruned_loss=0.05032, over 1438692.06 frames. ], batch size: 64, lr: 1.26e-02, grad_scale: 16.0 +2023-03-20 22:27:57,833 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.09 vs. limit=5.0 +2023-03-20 22:28:07,785 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.551e+02 2.283e+02 2.674e+02 3.489e+02 7.273e+02, threshold=5.348e+02, percent-clipped=5.0 +2023-03-20 22:28:16,904 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 22:28:17,428 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 22:28:20,443 INFO [train.py:901] (1/2) Epoch 13, batch 1150, loss[loss=0.1527, simple_loss=0.2205, pruned_loss=0.04246, over 7204.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2448, pruned_loss=0.04999, over 1438652.49 frames. ], batch size: 45, lr: 1.26e-02, grad_scale: 16.0 +2023-03-20 22:28:24,101 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8138, 2.1246, 1.6225, 2.8813, 2.5692, 2.9094, 2.1784, 2.1604], + device='cuda:1'), covar=tensor([0.2062, 0.0867, 0.3305, 0.0622, 0.0089, 0.0070, 0.0128, 0.0176], + device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0227, 0.0276, 0.0254, 0.0125, 0.0123, 0.0148, 0.0168], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:28:29,559 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 22:28:30,547 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 22:28:34,708 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35066.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 22:28:46,121 INFO [train.py:901] (1/2) Epoch 13, batch 1200, loss[loss=0.1863, simple_loss=0.257, pruned_loss=0.05785, over 7261.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2444, pruned_loss=0.04997, over 1441315.08 frames. ], batch size: 57, lr: 1.26e-02, grad_scale: 16.0 +2023-03-20 22:28:47,715 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35092.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:28:53,794 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8003, 3.8163, 3.1966, 3.1893, 3.2336, 2.1449, 1.4473, 3.8459], + device='cuda:1'), covar=tensor([0.0036, 0.0051, 0.0093, 0.0067, 0.0095, 0.0431, 0.0572, 0.0039], + device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0066, 0.0086, 0.0074, 0.0091, 0.0112, 0.0116, 0.0080], + device='cuda:1'), out_proj_covar=tensor([9.4520e-05, 9.5036e-05, 1.1464e-04, 1.0696e-04, 1.2211e-04, 1.5202e-04, + 1.5738e-04, 1.0616e-04], device='cuda:1') +2023-03-20 22:28:59,138 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.459e+02 2.344e+02 2.798e+02 3.632e+02 7.223e+02, threshold=5.596e+02, percent-clipped=4.0 +2023-03-20 22:29:03,161 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 22:29:11,746 INFO [train.py:901] (1/2) Epoch 13, batch 1250, loss[loss=0.183, simple_loss=0.262, pruned_loss=0.05197, over 7285.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2447, pruned_loss=0.05001, over 1440090.45 frames. ], batch size: 68, lr: 1.25e-02, grad_scale: 16.0 +2023-03-20 22:29:19,486 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35153.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:29:26,840 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 22:29:30,895 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 22:29:31,876 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 22:29:37,976 INFO [train.py:901] (1/2) Epoch 13, batch 1300, loss[loss=0.1621, simple_loss=0.2359, pruned_loss=0.04421, over 7349.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2461, pruned_loss=0.05061, over 1442656.23 frames. ], batch size: 73, lr: 1.25e-02, grad_scale: 16.0 +2023-03-20 22:29:45,078 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9271, 2.1454, 2.1215, 3.2833, 1.4567, 3.3728, 1.2587, 2.7701], + device='cuda:1'), covar=tensor([0.0052, 0.0952, 0.1558, 0.0066, 0.4032, 0.0075, 0.1084, 0.0165], + device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0261, 0.0293, 0.0148, 0.0286, 0.0152, 0.0264, 0.0207], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 22:29:50,990 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.414e+02 2.374e+02 2.864e+02 3.465e+02 9.066e+02, threshold=5.728e+02, percent-clipped=1.0 +2023-03-20 22:29:55,479 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 22:29:57,410 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 22:30:00,892 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 22:30:03,317 INFO [train.py:901] (1/2) Epoch 13, batch 1350, loss[loss=0.1814, simple_loss=0.2546, pruned_loss=0.05414, over 7343.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2462, pruned_loss=0.05058, over 1444945.76 frames. ], batch size: 63, lr: 1.25e-02, grad_scale: 16.0 +2023-03-20 22:30:08,890 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-20 22:30:09,030 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35249.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:30:12,417 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 22:30:23,997 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35278.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 22:30:25,976 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35282.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:30:29,429 INFO [train.py:901] (1/2) Epoch 13, batch 1400, loss[loss=0.1793, simple_loss=0.256, pruned_loss=0.05129, over 7357.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2461, pruned_loss=0.05043, over 1445543.19 frames. ], batch size: 63, lr: 1.25e-02, grad_scale: 16.0 +2023-03-20 22:30:33,587 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=35297.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:30:42,141 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.389e+02 2.860e+02 3.521e+02 6.843e+02, threshold=5.719e+02, percent-clipped=2.0 +2023-03-20 22:30:44,597 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 +2023-03-20 22:30:44,745 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 22:30:48,286 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=35326.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:30:51,011 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=35330.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:30:55,449 INFO [train.py:901] (1/2) Epoch 13, batch 1450, loss[loss=0.1744, simple_loss=0.2512, pruned_loss=0.04879, over 7318.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2456, pruned_loss=0.04981, over 1445540.31 frames. ], batch size: 59, lr: 1.25e-02, grad_scale: 16.0 +2023-03-20 22:30:58,629 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35345.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 22:31:09,109 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 22:31:09,733 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35366.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 22:31:13,728 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35374.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:31:21,007 INFO [train.py:901] (1/2) Epoch 13, batch 1500, loss[loss=0.15, simple_loss=0.2113, pruned_loss=0.04435, over 6234.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2454, pruned_loss=0.05007, over 1445308.42 frames. ], batch size: 27, lr: 1.25e-02, grad_scale: 16.0 +2023-03-20 22:31:24,918 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 22:31:29,930 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35406.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 22:31:31,595 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 22:31:33,721 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.695e+02 2.282e+02 2.834e+02 3.480e+02 6.162e+02, threshold=5.667e+02, percent-clipped=1.0 +2023-03-20 22:31:33,788 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=35414.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 22:31:45,069 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35435.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:31:46,861 INFO [train.py:901] (1/2) Epoch 13, batch 1550, loss[loss=0.1897, simple_loss=0.2596, pruned_loss=0.05992, over 7314.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2462, pruned_loss=0.0506, over 1446108.01 frames. ], batch size: 83, lr: 1.25e-02, grad_scale: 16.0 +2023-03-20 22:31:49,442 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 22:31:51,994 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35448.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:32:01,058 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.22 vs. limit=2.0 +2023-03-20 22:32:12,235 INFO [train.py:901] (1/2) Epoch 13, batch 1600, loss[loss=0.1679, simple_loss=0.242, pruned_loss=0.04693, over 7352.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2469, pruned_loss=0.05078, over 1446142.00 frames. ], batch size: 63, lr: 1.25e-02, grad_scale: 16.0 +2023-03-20 22:32:13,648 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-20 22:32:17,128 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 22:32:17,653 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 22:32:21,726 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 22:32:25,099 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.433e+02 2.407e+02 2.844e+02 3.510e+02 6.841e+02, threshold=5.688e+02, percent-clipped=1.0 +2023-03-20 22:32:32,149 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 22:32:33,791 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35530.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:32:36,209 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 22:32:38,167 INFO [train.py:901] (1/2) Epoch 13, batch 1650, loss[loss=0.1509, simple_loss=0.2318, pruned_loss=0.03501, over 7302.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2466, pruned_loss=0.0506, over 1444751.63 frames. ], batch size: 68, lr: 1.25e-02, grad_scale: 16.0 +2023-03-20 22:32:44,650 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 22:32:51,574 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.83 vs. limit=2.0 +2023-03-20 22:33:01,081 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2952, 1.4076, 1.3905, 1.5097, 1.5059, 1.2022, 1.4012, 0.8511], + device='cuda:1'), covar=tensor([0.0088, 0.0105, 0.0203, 0.0114, 0.0114, 0.0086, 0.0097, 0.0211], + device='cuda:1'), in_proj_covar=tensor([0.0023, 0.0021, 0.0021, 0.0020, 0.0023, 0.0020, 0.0021, 0.0027], + device='cuda:1'), out_proj_covar=tensor([2.6954e-05, 2.3887e-05, 2.5555e-05, 2.2336e-05, 2.7784e-05, 2.2627e-05, + 2.4113e-05, 3.3200e-05], device='cuda:1') +2023-03-20 22:33:02,021 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 22:33:03,518 INFO [train.py:901] (1/2) Epoch 13, batch 1700, loss[loss=0.1746, simple_loss=0.2475, pruned_loss=0.05084, over 7380.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2458, pruned_loss=0.05035, over 1444022.86 frames. ], batch size: 65, lr: 1.25e-02, grad_scale: 16.0 +2023-03-20 22:33:04,694 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35591.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:33:06,700 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 22:33:14,206 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.11 vs. limit=5.0 +2023-03-20 22:33:17,405 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.83 vs. limit=2.0 +2023-03-20 22:33:17,501 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.397e+02 2.386e+02 2.727e+02 3.355e+02 1.197e+03, threshold=5.455e+02, percent-clipped=3.0 +2023-03-20 22:33:18,503 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 22:33:29,974 INFO [train.py:901] (1/2) Epoch 13, batch 1750, loss[loss=0.1668, simple_loss=0.2457, pruned_loss=0.04401, over 7261.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2458, pruned_loss=0.05033, over 1443605.05 frames. ], batch size: 64, lr: 1.25e-02, grad_scale: 16.0 +2023-03-20 22:33:42,036 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 22:33:43,044 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 22:33:55,321 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5276, 2.8591, 2.1878, 2.7469, 2.6331, 2.1871, 2.7530, 2.5086], + device='cuda:1'), covar=tensor([0.0877, 0.0490, 0.1290, 0.0629, 0.1371, 0.0641, 0.1261, 0.1074], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0041, 0.0046, 0.0040, 0.0040, 0.0041, 0.0044, 0.0038], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 22:33:55,673 INFO [train.py:901] (1/2) Epoch 13, batch 1800, loss[loss=0.1726, simple_loss=0.2467, pruned_loss=0.04926, over 7271.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2459, pruned_loss=0.0504, over 1441424.33 frames. ], batch size: 70, lr: 1.25e-02, grad_scale: 16.0 +2023-03-20 22:34:02,310 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35701.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 22:34:04,892 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 22:34:09,416 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.652e+02 2.292e+02 2.730e+02 3.144e+02 6.344e+02, threshold=5.459e+02, percent-clipped=2.0 +2023-03-20 22:34:17,189 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35730.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:34:19,127 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 22:34:21,608 INFO [train.py:901] (1/2) Epoch 13, batch 1850, loss[loss=0.1455, simple_loss=0.2255, pruned_loss=0.03281, over 7317.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.245, pruned_loss=0.05006, over 1441239.67 frames. ], batch size: 44, lr: 1.24e-02, grad_scale: 16.0 +2023-03-20 22:34:26,208 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35748.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:34:29,144 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 22:34:38,456 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35771.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:34:46,014 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 22:34:47,982 INFO [train.py:901] (1/2) Epoch 13, batch 1900, loss[loss=0.1715, simple_loss=0.2395, pruned_loss=0.05172, over 7307.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.244, pruned_loss=0.0496, over 1441950.53 frames. ], batch size: 49, lr: 1.24e-02, grad_scale: 16.0 +2023-03-20 22:34:51,520 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=35796.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:34:52,251 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 +2023-03-20 22:34:59,520 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35811.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:35:01,329 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.430e+02 2.515e+02 2.934e+02 3.693e+02 9.378e+02, threshold=5.869e+02, percent-clipped=4.0 +2023-03-20 22:35:09,921 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 22:35:10,065 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35832.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:35:13,454 INFO [train.py:901] (1/2) Epoch 13, batch 1950, loss[loss=0.1642, simple_loss=0.2384, pruned_loss=0.04505, over 7218.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2442, pruned_loss=0.04957, over 1442703.02 frames. ], batch size: 93, lr: 1.24e-02, grad_scale: 16.0 +2023-03-20 22:35:14,989 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6632, 4.3204, 4.4138, 4.8581, 4.8480, 4.8039, 4.1503, 4.4047], + device='cuda:1'), covar=tensor([0.0782, 0.2259, 0.1962, 0.0905, 0.0600, 0.1184, 0.0883, 0.0894], + device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0291, 0.0235, 0.0221, 0.0171, 0.0286, 0.0158, 0.0203], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:35:20,962 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 22:35:25,843 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 22:35:26,362 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 22:35:26,454 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9295, 3.3133, 3.5643, 3.5170, 3.4285, 3.3977, 3.7329, 3.4809], + device='cuda:1'), covar=tensor([0.0106, 0.0208, 0.0132, 0.0181, 0.0348, 0.0136, 0.0162, 0.0127], + device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0069, 0.0068, 0.0058, 0.0112, 0.0073, 0.0070, 0.0069], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:35:31,043 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35872.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:35:37,872 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35886.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:35:39,342 INFO [train.py:901] (1/2) Epoch 13, batch 2000, loss[loss=0.154, simple_loss=0.2343, pruned_loss=0.03688, over 7363.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2443, pruned_loss=0.04937, over 1444800.41 frames. ], batch size: 73, lr: 1.24e-02, grad_scale: 16.0 +2023-03-20 22:35:43,310 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 22:35:52,259 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.519e+02 2.324e+02 2.736e+02 3.449e+02 6.925e+02, threshold=5.472e+02, percent-clipped=1.0 +2023-03-20 22:35:54,296 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 22:36:02,377 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 22:36:05,051 INFO [train.py:901] (1/2) Epoch 13, batch 2050, loss[loss=0.1854, simple_loss=0.2596, pruned_loss=0.05559, over 7293.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2448, pruned_loss=0.04958, over 1446325.12 frames. ], batch size: 86, lr: 1.24e-02, grad_scale: 16.0 +2023-03-20 22:36:12,568 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3594, 2.3779, 1.9158, 3.0969, 1.5746, 3.4001, 1.4502, 3.0655], + device='cuda:1'), covar=tensor([0.0079, 0.0759, 0.1838, 0.0051, 0.4049, 0.0110, 0.1014, 0.0281], + device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0267, 0.0302, 0.0150, 0.0290, 0.0160, 0.0270, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 22:36:15,766 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-20 22:36:24,217 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4372, 2.9087, 2.1553, 2.5525, 2.5473, 2.0833, 2.9044, 2.4392], + device='cuda:1'), covar=tensor([0.1109, 0.0622, 0.1368, 0.1855, 0.1510, 0.1076, 0.1515, 0.1415], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0040, 0.0047, 0.0041, 0.0041, 0.0041, 0.0044, 0.0038], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 22:36:30,611 INFO [train.py:901] (1/2) Epoch 13, batch 2100, loss[loss=0.18, simple_loss=0.2571, pruned_loss=0.05145, over 7318.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.244, pruned_loss=0.0494, over 1444475.97 frames. ], batch size: 49, lr: 1.24e-02, grad_scale: 16.0 +2023-03-20 22:36:33,748 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0506, 2.3712, 1.8694, 2.7816, 2.6296, 2.4134, 2.3979, 2.2095], + device='cuda:1'), covar=tensor([0.1762, 0.0850, 0.2973, 0.0409, 0.0071, 0.0046, 0.0127, 0.0222], + device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0229, 0.0268, 0.0251, 0.0125, 0.0122, 0.0147, 0.0168], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:36:34,214 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1995, 1.0535, 1.4799, 1.5780, 1.4145, 1.3658, 1.4854, 1.3840], + device='cuda:1'), covar=tensor([0.1609, 0.1899, 0.1061, 0.1460, 0.2297, 0.2059, 0.1605, 0.4471], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0049, 0.0035, 0.0033, 0.0041, 0.0040, 0.0045, 0.0039], + device='cuda:1'), out_proj_covar=tensor([1.0089e-04, 1.1762e-04, 8.9117e-05, 8.8140e-05, 1.0222e-04, 1.0295e-04, + 1.1289e-04, 1.0211e-04], device='cuda:1') +2023-03-20 22:36:40,589 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36001.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 22:36:41,448 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 22:36:43,923 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 22:36:47,331 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.503e+02 2.959e+02 3.562e+02 9.049e+02, threshold=5.919e+02, percent-clipped=3.0 +2023-03-20 22:36:55,477 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36030.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:36:59,848 INFO [train.py:901] (1/2) Epoch 13, batch 2150, loss[loss=0.1654, simple_loss=0.2388, pruned_loss=0.04599, over 7332.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.244, pruned_loss=0.04943, over 1442775.76 frames. ], batch size: 59, lr: 1.24e-02, grad_scale: 16.0 +2023-03-20 22:37:05,369 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=36049.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 22:37:10,916 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.9824, 4.5630, 4.5405, 5.1390, 4.9763, 5.1230, 4.6560, 4.6538], + device='cuda:1'), covar=tensor([0.0782, 0.2773, 0.2479, 0.1308, 0.0909, 0.1283, 0.0741, 0.0923], + device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0304, 0.0248, 0.0238, 0.0177, 0.0297, 0.0163, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:37:20,002 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=36078.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:37:25,436 INFO [train.py:901] (1/2) Epoch 13, batch 2200, loss[loss=0.2286, simple_loss=0.2893, pruned_loss=0.08394, over 7314.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.245, pruned_loss=0.05014, over 1443292.51 frames. ], batch size: 59, lr: 1.24e-02, grad_scale: 16.0 +2023-03-20 22:37:27,443 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 22:37:39,137 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.739e+02 2.623e+02 3.188e+02 4.090e+02 7.570e+02, threshold=6.377e+02, percent-clipped=4.0 +2023-03-20 22:37:45,247 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36127.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:37:51,825 INFO [train.py:901] (1/2) Epoch 13, batch 2250, loss[loss=0.1244, simple_loss=0.1858, pruned_loss=0.03148, over 6054.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2441, pruned_loss=0.04943, over 1442271.57 frames. ], batch size: 26, lr: 1.24e-02, grad_scale: 16.0 +2023-03-20 22:37:55,594 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.49 vs. limit=5.0 +2023-03-20 22:38:01,414 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 22:38:01,426 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 22:38:05,929 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36167.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:38:07,500 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0344, 0.8666, 1.3712, 1.3791, 1.3314, 1.3564, 1.1766, 1.3310], + device='cuda:1'), covar=tensor([0.1581, 0.3006, 0.0769, 0.1186, 0.2110, 0.1090, 0.1095, 0.2998], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0051, 0.0035, 0.0033, 0.0041, 0.0040, 0.0046, 0.0040], + device='cuda:1'), out_proj_covar=tensor([1.0293e-04, 1.2108e-04, 9.0504e-05, 9.0155e-05, 1.0305e-04, 1.0376e-04, + 1.1488e-04, 1.0331e-04], device='cuda:1') +2023-03-20 22:38:12,975 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36181.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:38:13,832 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 22:38:15,436 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36186.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:38:16,803 INFO [train.py:901] (1/2) Epoch 13, batch 2300, loss[loss=0.1733, simple_loss=0.2434, pruned_loss=0.0516, over 7370.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2449, pruned_loss=0.04962, over 1441743.18 frames. ], batch size: 51, lr: 1.24e-02, grad_scale: 16.0 +2023-03-20 22:38:30,767 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.358e+02 2.395e+02 2.775e+02 3.184e+02 6.455e+02, threshold=5.551e+02, percent-clipped=1.0 +2023-03-20 22:38:38,288 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 +2023-03-20 22:38:40,906 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=36234.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:38:40,943 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6198, 5.0585, 5.1298, 5.0731, 4.8617, 4.6838, 5.1616, 4.9377], + device='cuda:1'), covar=tensor([0.0383, 0.0422, 0.0468, 0.0459, 0.0320, 0.0291, 0.0361, 0.0500], + device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0191, 0.0141, 0.0139, 0.0118, 0.0176, 0.0150, 0.0119], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:38:43,002 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5368, 3.8980, 4.1192, 4.2399, 4.0341, 4.2145, 4.4290, 4.0613], + device='cuda:1'), covar=tensor([0.0135, 0.0132, 0.0113, 0.0108, 0.0321, 0.0085, 0.0135, 0.0113], + device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0066, 0.0067, 0.0057, 0.0111, 0.0073, 0.0070, 0.0068], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:38:43,399 INFO [train.py:901] (1/2) Epoch 13, batch 2350, loss[loss=0.1453, simple_loss=0.2132, pruned_loss=0.03868, over 7171.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2449, pruned_loss=0.04947, over 1441726.34 frames. ], batch size: 39, lr: 1.24e-02, grad_scale: 16.0 +2023-03-20 22:38:45,089 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36242.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:38:51,293 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-03-20 22:38:59,478 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 22:39:06,135 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 22:39:09,119 INFO [train.py:901] (1/2) Epoch 13, batch 2400, loss[loss=0.1632, simple_loss=0.2409, pruned_loss=0.04272, over 7344.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2442, pruned_loss=0.04893, over 1442394.09 frames. ], batch size: 73, lr: 1.24e-02, grad_scale: 16.0 +2023-03-20 22:39:18,270 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 22:39:20,706 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 22:39:22,694 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.408e+02 2.253e+02 2.752e+02 3.236e+02 5.663e+02, threshold=5.504e+02, percent-clipped=1.0 +2023-03-20 22:39:26,037 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 +2023-03-20 22:39:34,626 INFO [train.py:901] (1/2) Epoch 13, batch 2450, loss[loss=0.1741, simple_loss=0.2467, pruned_loss=0.05072, over 7359.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2444, pruned_loss=0.04922, over 1444809.08 frames. ], batch size: 51, lr: 1.23e-02, grad_scale: 16.0 +2023-03-20 22:39:45,329 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 22:39:51,987 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4983, 5.0602, 5.2098, 5.0405, 4.7836, 4.6322, 5.1839, 4.8708], + device='cuda:1'), covar=tensor([0.0400, 0.0380, 0.0278, 0.0471, 0.0350, 0.0314, 0.0245, 0.0520], + device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0187, 0.0136, 0.0137, 0.0114, 0.0172, 0.0145, 0.0114], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:40:00,886 INFO [train.py:901] (1/2) Epoch 13, batch 2500, loss[loss=0.1663, simple_loss=0.2409, pruned_loss=0.04585, over 7343.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2443, pruned_loss=0.04946, over 1443614.12 frames. ], batch size: 61, lr: 1.23e-02, grad_scale: 16.0 +2023-03-20 22:40:02,730 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-20 22:40:11,280 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 22:40:14,272 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.589e+02 2.433e+02 2.922e+02 3.588e+02 6.590e+02, threshold=5.845e+02, percent-clipped=3.0 +2023-03-20 22:40:20,420 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36427.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:40:26,226 INFO [train.py:901] (1/2) Epoch 13, batch 2550, loss[loss=0.1795, simple_loss=0.2576, pruned_loss=0.05069, over 7127.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2433, pruned_loss=0.04926, over 1437762.35 frames. ], batch size: 98, lr: 1.23e-02, grad_scale: 16.0 +2023-03-20 22:40:37,481 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1494, 1.3753, 1.1527, 1.2944, 1.1323, 1.0351, 0.9500, 0.7553], + device='cuda:1'), covar=tensor([0.0112, 0.0103, 0.0224, 0.0094, 0.0202, 0.0105, 0.0117, 0.0181], + device='cuda:1'), in_proj_covar=tensor([0.0022, 0.0019, 0.0020, 0.0020, 0.0022, 0.0020, 0.0020, 0.0027], + device='cuda:1'), out_proj_covar=tensor([2.5952e-05, 2.2366e-05, 2.3775e-05, 2.2409e-05, 2.6427e-05, 2.2092e-05, + 2.3446e-05, 3.2348e-05], device='cuda:1') +2023-03-20 22:40:40,950 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36467.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:40:45,495 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=36475.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:40:52,334 INFO [train.py:901] (1/2) Epoch 13, batch 2600, loss[loss=0.1737, simple_loss=0.245, pruned_loss=0.05119, over 7273.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2434, pruned_loss=0.04912, over 1439245.30 frames. ], batch size: 47, lr: 1.23e-02, grad_scale: 16.0 +2023-03-20 22:40:56,638 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.07 vs. limit=5.0 +2023-03-20 22:41:01,695 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4338, 3.7765, 3.5662, 3.7291, 3.4991, 3.5584, 3.7974, 3.9580], + device='cuda:1'), covar=tensor([0.0338, 0.0190, 0.0272, 0.0307, 0.0406, 0.0468, 0.0364, 0.0281], + device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0101, 0.0092, 0.0102, 0.0092, 0.0080, 0.0075, 0.0079], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:41:05,000 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.441e+02 2.354e+02 2.833e+02 3.517e+02 6.810e+02, threshold=5.665e+02, percent-clipped=1.0 +2023-03-20 22:41:05,076 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=36515.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:41:07,923 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2536, 4.7880, 4.8864, 4.8210, 4.7339, 4.3026, 4.9484, 4.7516], + device='cuda:1'), covar=tensor([0.0419, 0.0396, 0.0386, 0.0399, 0.0294, 0.0336, 0.0286, 0.0490], + device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0188, 0.0138, 0.0135, 0.0115, 0.0173, 0.0146, 0.0115], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:41:15,737 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36537.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:41:16,681 INFO [train.py:901] (1/2) Epoch 13, batch 2650, loss[loss=0.1527, simple_loss=0.2298, pruned_loss=0.03775, over 7344.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2424, pruned_loss=0.0487, over 1439826.39 frames. ], batch size: 44, lr: 1.23e-02, grad_scale: 16.0 +2023-03-20 22:41:25,036 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-20 22:41:41,360 INFO [train.py:901] (1/2) Epoch 13, batch 2700, loss[loss=0.1418, simple_loss=0.1922, pruned_loss=0.04572, over 6192.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2424, pruned_loss=0.04865, over 1440157.49 frames. ], batch size: 26, lr: 1.23e-02, grad_scale: 16.0 +2023-03-20 22:41:42,400 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2788, 3.7465, 3.8788, 3.9329, 3.7255, 3.8351, 4.1204, 3.4537], + device='cuda:1'), covar=tensor([0.0083, 0.0135, 0.0107, 0.0119, 0.0301, 0.0107, 0.0133, 0.0167], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0069, 0.0068, 0.0058, 0.0114, 0.0076, 0.0072, 0.0072], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:41:49,627 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36605.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:41:50,835 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 +2023-03-20 22:41:51,131 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36608.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:41:53,058 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36612.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:41:54,401 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.842e+02 2.424e+02 2.835e+02 3.228e+02 5.690e+02, threshold=5.669e+02, percent-clipped=1.0 +2023-03-20 22:42:06,222 INFO [train.py:901] (1/2) Epoch 13, batch 2750, loss[loss=0.1819, simple_loss=0.2547, pruned_loss=0.05451, over 7313.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2427, pruned_loss=0.04908, over 1439815.79 frames. ], batch size: 49, lr: 1.23e-02, grad_scale: 16.0 +2023-03-20 22:42:08,925 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 22:42:13,936 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7046, 4.6331, 4.3918, 3.8915, 4.1952, 2.9060, 2.5163, 4.6754], + device='cuda:1'), covar=tensor([0.0021, 0.0032, 0.0042, 0.0043, 0.0042, 0.0310, 0.0411, 0.0032], + device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0064, 0.0085, 0.0070, 0.0089, 0.0110, 0.0113, 0.0077], + device='cuda:1'), out_proj_covar=tensor([9.3034e-05, 9.3753e-05, 1.1236e-04, 1.0008e-04, 1.1943e-04, 1.4857e-04, + 1.5244e-04, 1.0160e-04], device='cuda:1') +2023-03-20 22:42:19,330 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36666.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:42:20,749 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36669.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:42:23,002 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36673.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:42:29,776 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6522, 3.5033, 3.2621, 3.4190, 2.7722, 2.5465, 3.6058, 2.6400], + device='cuda:1'), covar=tensor([0.0183, 0.0221, 0.0222, 0.0225, 0.0274, 0.0374, 0.0288, 0.0709], + device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0292, 0.0241, 0.0302, 0.0300, 0.0294, 0.0290, 0.0289], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 22:42:30,537 INFO [train.py:901] (1/2) Epoch 13, batch 2800, loss[loss=0.1711, simple_loss=0.2509, pruned_loss=0.04567, over 7328.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2436, pruned_loss=0.04957, over 1442813.70 frames. ], batch size: 61, lr: 1.23e-02, grad_scale: 16.0 +2023-03-20 22:42:32,093 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3144, 3.7131, 3.7350, 3.8643, 3.6730, 3.7076, 4.1103, 3.3944], + device='cuda:1'), covar=tensor([0.0092, 0.0168, 0.0145, 0.0141, 0.0363, 0.0106, 0.0140, 0.0192], + device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0070, 0.0071, 0.0060, 0.0118, 0.0078, 0.0074, 0.0074], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:42:35,987 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36700.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 22:42:40,317 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2101, 4.6589, 4.2498, 4.6063, 4.3072, 4.5022, 4.7940, 4.8766], + device='cuda:1'), covar=tensor([0.0188, 0.0108, 0.0160, 0.0126, 0.0264, 0.0151, 0.0153, 0.0113], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0102, 0.0094, 0.0103, 0.0093, 0.0083, 0.0078, 0.0079], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:42:40,623 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 +2023-03-20 22:42:56,462 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 22:42:56,512 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 22:42:56,537 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 22:42:56,552 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 22:42:57,049 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 22:42:57,064 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 22:42:57,075 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 22:42:57,171 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 22:42:57,245 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 22:42:57,261 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 22:42:57,277 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 22:42:57,435 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 22:42:57,440 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 22:42:57,448 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 22:42:57,585 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 22:42:57,588 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 22:42:57,645 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 22:43:05,226 INFO [train.py:901] (1/2) Epoch 14, batch 0, loss[loss=0.1901, simple_loss=0.265, pruned_loss=0.05761, over 7349.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.265, pruned_loss=0.05761, over 7349.00 frames. ], batch size: 73, lr: 1.19e-02, grad_scale: 16.0 +2023-03-20 22:43:05,226 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 22:43:14,994 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1764, 1.3613, 1.4879, 1.4715, 1.5665, 1.5973, 1.4540, 1.4304], + device='cuda:1'), covar=tensor([0.1279, 0.1363, 0.0666, 0.0523, 0.1413, 0.1816, 0.1032, 0.2137], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0045, 0.0033, 0.0030, 0.0037, 0.0037, 0.0043, 0.0036], + device='cuda:1'), out_proj_covar=tensor([9.5259e-05, 1.0952e-04, 8.3993e-05, 8.3061e-05, 9.5528e-05, 9.7744e-05, + 1.0739e-04, 9.5338e-05], device='cuda:1') +2023-03-20 22:43:15,695 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6903, 4.2030, 4.2603, 4.2834, 4.3581, 4.0276, 4.4321, 3.9941], + device='cuda:1'), covar=tensor([0.0051, 0.0132, 0.0096, 0.0113, 0.0224, 0.0104, 0.0111, 0.0144], + device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0070, 0.0071, 0.0060, 0.0119, 0.0079, 0.0074, 0.0074], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:43:15,953 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5788, 2.9024, 1.7634, 3.7005, 2.1783, 2.8441, 1.4714, 1.7844], + device='cuda:1'), covar=tensor([0.0197, 0.0429, 0.1752, 0.0353, 0.0275, 0.0415, 0.2656, 0.1401], + device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0226, 0.0299, 0.0218, 0.0238, 0.0223, 0.0262, 0.0280], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 22:43:18,034 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6603, 4.8659, 4.7710, 4.8765, 4.4675, 4.3910, 4.8913, 4.4717], + device='cuda:1'), covar=tensor([0.0323, 0.0405, 0.0466, 0.0393, 0.0542, 0.0395, 0.0319, 0.0685], + device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0191, 0.0140, 0.0137, 0.0118, 0.0175, 0.0147, 0.0116], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:43:21,194 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1729, 3.1285, 3.5286, 3.2754, 3.4278, 3.0518, 2.5255, 3.2062], + device='cuda:1'), covar=tensor([0.1735, 0.0718, 0.0818, 0.1395, 0.0903, 0.1473, 0.3201, 0.1900], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0040, 0.0035, 0.0037, 0.0034, 0.0032, 0.0049, 0.0037], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:43:31,788 INFO [train.py:935] (1/2) Epoch 14, validation: loss=0.1698, simple_loss=0.2569, pruned_loss=0.04131, over 1622729.00 frames. +2023-03-20 22:43:31,789 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-20 22:43:32,794 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 2.144e+02 2.550e+02 3.051e+02 8.229e+02, threshold=5.101e+02, percent-clipped=3.0 +2023-03-20 22:43:38,811 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 22:43:50,997 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 22:43:56,593 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36761.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 22:43:57,422 INFO [train.py:901] (1/2) Epoch 14, batch 50, loss[loss=0.174, simple_loss=0.251, pruned_loss=0.04852, over 7271.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2439, pruned_loss=0.05108, over 326105.95 frames. ], batch size: 57, lr: 1.19e-02, grad_scale: 16.0 +2023-03-20 22:43:57,441 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 22:44:00,037 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 22:44:02,581 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 22:44:12,281 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2592, 2.2920, 2.2097, 3.6104, 1.4511, 3.3007, 1.5757, 2.6927], + device='cuda:1'), covar=tensor([0.0061, 0.1024, 0.1577, 0.0063, 0.3921, 0.0092, 0.0971, 0.0158], + device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0259, 0.0300, 0.0148, 0.0288, 0.0158, 0.0267, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 22:44:19,185 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36805.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 22:44:21,219 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 22:44:21,742 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 22:44:23,687 INFO [train.py:901] (1/2) Epoch 14, batch 100, loss[loss=0.1456, simple_loss=0.2232, pruned_loss=0.03398, over 7320.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2435, pruned_loss=0.04935, over 575140.59 frames. ], batch size: 44, lr: 1.19e-02, grad_scale: 16.0 +2023-03-20 22:44:24,632 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.278e+02 2.289e+02 2.637e+02 3.349e+02 8.885e+02, threshold=5.275e+02, percent-clipped=4.0 +2023-03-20 22:44:36,380 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36837.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:44:49,234 INFO [train.py:901] (1/2) Epoch 14, batch 150, loss[loss=0.2205, simple_loss=0.2746, pruned_loss=0.08325, over 7154.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2457, pruned_loss=0.04952, over 768987.92 frames. ], batch size: 98, lr: 1.19e-02, grad_scale: 16.0 +2023-03-20 22:44:50,850 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36866.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 22:44:52,916 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9468, 2.1808, 1.7749, 3.1357, 2.8741, 3.0019, 1.8520, 2.0425], + device='cuda:1'), covar=tensor([0.1673, 0.0719, 0.2910, 0.0423, 0.0068, 0.0073, 0.0092, 0.0116], + device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0229, 0.0273, 0.0255, 0.0129, 0.0123, 0.0149, 0.0169], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:45:00,635 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=36885.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:45:14,404 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36911.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:45:15,261 INFO [train.py:901] (1/2) Epoch 14, batch 200, loss[loss=0.1742, simple_loss=0.2428, pruned_loss=0.05282, over 7250.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2433, pruned_loss=0.04846, over 916908.63 frames. ], batch size: 55, lr: 1.19e-02, grad_scale: 16.0 +2023-03-20 22:45:16,281 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 2.238e+02 2.665e+02 3.334e+02 5.724e+02, threshold=5.331e+02, percent-clipped=5.0 +2023-03-20 22:45:21,457 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 22:45:26,019 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 22:45:32,051 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 22:45:32,459 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 +2023-03-20 22:45:38,677 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36958.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:45:40,123 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36961.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:45:41,022 INFO [train.py:901] (1/2) Epoch 14, batch 250, loss[loss=0.1693, simple_loss=0.2476, pruned_loss=0.04551, over 7271.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2432, pruned_loss=0.04864, over 1032034.48 frames. ], batch size: 77, lr: 1.19e-02, grad_scale: 16.0 +2023-03-20 22:45:41,609 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36964.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:45:43,611 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36968.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:45:44,550 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 22:45:45,691 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36972.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:45:54,313 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6878, 3.2735, 3.6136, 3.7617, 3.7083, 3.7993, 3.5194, 3.5688], + device='cuda:1'), covar=tensor([0.0029, 0.0087, 0.0031, 0.0035, 0.0029, 0.0028, 0.0050, 0.0044], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0046, 0.0040, 0.0038, 0.0039, 0.0041, 0.0045, 0.0048], + device='cuda:1'), out_proj_covar=tensor([7.8852e-05, 1.2538e-04, 1.0614e-04, 9.1085e-05, 9.4487e-05, 1.0171e-04, + 1.1884e-04, 1.2109e-04], device='cuda:1') +2023-03-20 22:46:03,274 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37004.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:46:05,760 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 22:46:07,732 INFO [train.py:901] (1/2) Epoch 14, batch 300, loss[loss=0.166, simple_loss=0.2418, pruned_loss=0.04506, over 7357.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.244, pruned_loss=0.04908, over 1124395.34 frames. ], batch size: 63, lr: 1.18e-02, grad_scale: 16.0 +2023-03-20 22:46:08,577 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 +2023-03-20 22:46:08,723 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.157e+02 2.665e+02 3.235e+02 4.980e+02, threshold=5.329e+02, percent-clipped=0.0 +2023-03-20 22:46:10,891 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37019.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:46:13,968 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6263, 2.7528, 1.6366, 3.2295, 1.9107, 2.5375, 1.4104, 1.5795], + device='cuda:1'), covar=tensor([0.0192, 0.0464, 0.1872, 0.0369, 0.0285, 0.0333, 0.2698, 0.1641], + device='cuda:1'), in_proj_covar=tensor([0.0218, 0.0228, 0.0303, 0.0217, 0.0244, 0.0227, 0.0264, 0.0284], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-20 22:46:14,804 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 22:46:19,033 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 +2023-03-20 22:46:26,511 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-20 22:46:29,236 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37056.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 22:46:32,658 INFO [train.py:901] (1/2) Epoch 14, batch 350, loss[loss=0.1471, simple_loss=0.2335, pruned_loss=0.03041, over 7326.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2432, pruned_loss=0.04871, over 1195222.42 frames. ], batch size: 83, lr: 1.18e-02, grad_scale: 16.0 +2023-03-20 22:46:34,442 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37065.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:46:48,907 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 22:46:58,915 INFO [train.py:901] (1/2) Epoch 14, batch 400, loss[loss=0.2151, simple_loss=0.2804, pruned_loss=0.07495, over 7208.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2437, pruned_loss=0.04906, over 1251382.81 frames. ], batch size: 93, lr: 1.18e-02, grad_scale: 16.0 +2023-03-20 22:46:59,939 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.519e+02 2.445e+02 2.867e+02 4.022e+02 1.109e+03, threshold=5.735e+02, percent-clipped=7.0 +2023-03-20 22:47:14,283 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0089, 2.3715, 3.1151, 2.8736, 3.0201, 2.5314, 2.4762, 2.7040], + device='cuda:1'), covar=tensor([0.1192, 0.0891, 0.1052, 0.2240, 0.1230, 0.1717, 0.1735, 0.2033], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0041, 0.0036, 0.0037, 0.0034, 0.0032, 0.0049, 0.0037], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:47:15,935 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-20 22:47:23,796 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37161.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 22:47:24,742 INFO [train.py:901] (1/2) Epoch 14, batch 450, loss[loss=0.1373, simple_loss=0.2096, pruned_loss=0.03247, over 7334.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2431, pruned_loss=0.04809, over 1293911.44 frames. ], batch size: 44, lr: 1.18e-02, grad_scale: 16.0 +2023-03-20 22:47:27,182 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 +2023-03-20 22:47:27,991 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1178, 3.2025, 2.0358, 3.7730, 2.8814, 3.0977, 1.6297, 1.9040], + device='cuda:1'), covar=tensor([0.0176, 0.0497, 0.1834, 0.0360, 0.0253, 0.0303, 0.2454, 0.1731], + device='cuda:1'), in_proj_covar=tensor([0.0214, 0.0223, 0.0296, 0.0213, 0.0240, 0.0225, 0.0260, 0.0280], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 22:47:30,311 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 22:47:30,324 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 22:47:36,315 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-20 22:47:50,857 INFO [train.py:901] (1/2) Epoch 14, batch 500, loss[loss=0.1737, simple_loss=0.2462, pruned_loss=0.05058, over 7305.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2424, pruned_loss=0.04798, over 1324305.29 frames. ], batch size: 80, lr: 1.18e-02, grad_scale: 16.0 +2023-03-20 22:47:51,840 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.393e+02 2.247e+02 2.615e+02 3.199e+02 8.743e+02, threshold=5.229e+02, percent-clipped=2.0 +2023-03-20 22:48:02,386 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 22:48:03,345 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 22:48:04,344 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 22:48:06,864 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 22:48:09,530 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6437, 2.7404, 2.5035, 3.7409, 1.5714, 3.4934, 1.7334, 3.0343], + device='cuda:1'), covar=tensor([0.0064, 0.0778, 0.1486, 0.0045, 0.4072, 0.0082, 0.0931, 0.0140], + device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0258, 0.0297, 0.0146, 0.0284, 0.0158, 0.0265, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 22:48:10,293 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 +2023-03-20 22:48:11,382 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 22:48:16,019 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37261.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:48:16,926 INFO [train.py:901] (1/2) Epoch 14, batch 550, loss[loss=0.1519, simple_loss=0.2191, pruned_loss=0.04232, over 7028.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2418, pruned_loss=0.0479, over 1351356.22 frames. ], batch size: 35, lr: 1.18e-02, grad_scale: 16.0 +2023-03-20 22:48:17,556 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37264.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:48:18,974 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37267.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:48:19,488 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:48:19,507 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:48:22,841 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 22:48:30,149 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-03-20 22:48:30,366 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 22:48:33,773 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 22:48:39,728 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=37309.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:48:41,178 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 22:48:41,224 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=37312.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:48:41,660 INFO [train.py:901] (1/2) Epoch 14, batch 600, loss[loss=0.169, simple_loss=0.2401, pruned_loss=0.04897, over 7272.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2426, pruned_loss=0.04842, over 1372403.04 frames. ], batch size: 77, lr: 1.18e-02, grad_scale: 16.0 +2023-03-20 22:48:42,264 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37314.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:48:42,683 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.746e+02 2.422e+02 2.800e+02 3.690e+02 8.358e+02, threshold=5.600e+02, percent-clipped=9.0 +2023-03-20 22:48:43,316 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=37316.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:48:50,502 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37329.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:48:57,804 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 22:49:04,304 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37356.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 22:49:06,310 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37360.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:49:06,747 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 22:49:07,725 INFO [train.py:901] (1/2) Epoch 14, batch 650, loss[loss=0.1555, simple_loss=0.24, pruned_loss=0.03552, over 7197.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2425, pruned_loss=0.04849, over 1386244.11 frames. ], batch size: 50, lr: 1.18e-02, grad_scale: 16.0 +2023-03-20 22:49:08,328 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37364.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:49:23,354 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6821, 2.1426, 2.7249, 2.5816, 2.7411, 2.4398, 2.0653, 2.6392], + device='cuda:1'), covar=tensor([0.0873, 0.1219, 0.1042, 0.1255, 0.0579, 0.0803, 0.2465, 0.1106], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0041, 0.0035, 0.0037, 0.0033, 0.0032, 0.0048, 0.0037], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:49:23,717 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 22:49:27,977 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7174, 4.3358, 4.1603, 4.6379, 4.6148, 4.7167, 3.9498, 4.2397], + device='cuda:1'), covar=tensor([0.0610, 0.2200, 0.1822, 0.0987, 0.0688, 0.1121, 0.0744, 0.1016], + device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0296, 0.0237, 0.0225, 0.0172, 0.0287, 0.0160, 0.0204], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:49:28,483 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=37404.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 22:49:29,545 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6898, 3.9143, 3.6962, 3.9450, 3.7351, 3.9155, 4.1733, 4.2652], + device='cuda:1'), covar=tensor([0.0228, 0.0169, 0.0206, 0.0167, 0.0308, 0.0266, 0.0275, 0.0202], + device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0103, 0.0096, 0.0107, 0.0096, 0.0086, 0.0081, 0.0080], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:49:32,038 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 22:49:33,568 INFO [train.py:901] (1/2) Epoch 14, batch 700, loss[loss=0.1659, simple_loss=0.2468, pruned_loss=0.04251, over 7133.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2436, pruned_loss=0.04861, over 1399601.82 frames. ], batch size: 98, lr: 1.18e-02, grad_scale: 16.0 +2023-03-20 22:49:34,576 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.428e+02 2.290e+02 2.741e+02 3.295e+02 7.204e+02, threshold=5.481e+02, percent-clipped=3.0 +2023-03-20 22:49:37,368 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.26 vs. limit=5.0 +2023-03-20 22:49:39,687 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37425.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:49:47,694 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0526, 4.4905, 4.4449, 4.9191, 4.9108, 4.9531, 4.1637, 4.4519], + device='cuda:1'), covar=tensor([0.0630, 0.2140, 0.2233, 0.1019, 0.0834, 0.1102, 0.0833, 0.0938], + device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0298, 0.0240, 0.0228, 0.0175, 0.0293, 0.0163, 0.0207], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:49:55,076 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 22:49:55,603 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 22:49:58,160 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37461.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 22:49:59,026 INFO [train.py:901] (1/2) Epoch 14, batch 750, loss[loss=0.1672, simple_loss=0.2453, pruned_loss=0.04456, over 7347.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2435, pruned_loss=0.04836, over 1411658.69 frames. ], batch size: 75, lr: 1.18e-02, grad_scale: 16.0 +2023-03-20 22:50:09,051 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 22:50:14,008 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 22:50:21,195 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 22:50:22,196 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 22:50:22,736 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=37509.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 22:50:24,694 INFO [train.py:901] (1/2) Epoch 14, batch 800, loss[loss=0.1693, simple_loss=0.2489, pruned_loss=0.04488, over 7294.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2429, pruned_loss=0.04847, over 1416577.86 frames. ], batch size: 68, lr: 1.18e-02, grad_scale: 16.0 +2023-03-20 22:50:25,667 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.705e+02 2.326e+02 2.652e+02 3.286e+02 4.306e+02, threshold=5.303e+02, percent-clipped=0.0 +2023-03-20 22:50:33,735 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 22:50:50,151 INFO [train.py:901] (1/2) Epoch 14, batch 850, loss[loss=0.1511, simple_loss=0.2142, pruned_loss=0.04397, over 7039.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2423, pruned_loss=0.04837, over 1422866.20 frames. ], batch size: 35, lr: 1.18e-02, grad_scale: 16.0 +2023-03-20 22:50:52,142 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 22:50:52,148 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 22:50:52,239 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37567.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:50:57,221 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 22:51:00,886 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 22:51:16,871 INFO [train.py:901] (1/2) Epoch 14, batch 900, loss[loss=0.1291, simple_loss=0.212, pruned_loss=0.02314, over 7131.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2426, pruned_loss=0.04832, over 1426816.09 frames. ], batch size: 41, lr: 1.17e-02, grad_scale: 16.0 +2023-03-20 22:51:17,481 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37614.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:51:17,855 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.554e+02 2.227e+02 2.893e+02 3.538e+02 7.539e+02, threshold=5.787e+02, percent-clipped=4.0 +2023-03-20 22:51:17,928 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=37615.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:51:22,544 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37624.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:51:32,006 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5785, 4.9763, 4.9682, 4.9583, 4.8283, 4.5274, 5.0501, 4.8475], + device='cuda:1'), covar=tensor([0.0367, 0.0382, 0.0411, 0.0476, 0.0353, 0.0346, 0.0364, 0.0509], + device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0194, 0.0139, 0.0141, 0.0121, 0.0177, 0.0153, 0.0119], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:51:37,925 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 22:51:40,453 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37660.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:51:41,339 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=37662.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:51:41,776 INFO [train.py:901] (1/2) Epoch 14, batch 950, loss[loss=0.177, simple_loss=0.2449, pruned_loss=0.05453, over 7233.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2425, pruned_loss=0.04796, over 1432022.41 frames. ], batch size: 45, lr: 1.17e-02, grad_scale: 16.0 +2023-03-20 22:51:50,858 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 +2023-03-20 22:52:00,926 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6475, 3.3642, 3.2893, 3.2452, 2.8582, 2.7163, 3.4023, 2.7030], + device='cuda:1'), covar=tensor([0.0208, 0.0227, 0.0209, 0.0207, 0.0304, 0.0438, 0.0268, 0.0719], + device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0293, 0.0245, 0.0310, 0.0297, 0.0298, 0.0294, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 22:52:01,397 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0752, 1.5853, 1.4572, 1.2944, 1.2780, 1.1565, 0.9385, 0.9196], + device='cuda:1'), covar=tensor([0.0129, 0.0070, 0.0183, 0.0106, 0.0172, 0.0135, 0.0102, 0.0121], + device='cuda:1'), in_proj_covar=tensor([0.0022, 0.0020, 0.0020, 0.0020, 0.0022, 0.0020, 0.0021, 0.0027], + device='cuda:1'), out_proj_covar=tensor([2.6727e-05, 2.3039e-05, 2.4310e-05, 2.2006e-05, 2.7378e-05, 2.2733e-05, + 2.4161e-05, 3.2063e-05], device='cuda:1') +2023-03-20 22:52:03,236 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 22:52:05,809 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=37708.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:52:08,356 INFO [train.py:901] (1/2) Epoch 14, batch 1000, loss[loss=0.146, simple_loss=0.2184, pruned_loss=0.03675, over 7256.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2418, pruned_loss=0.04763, over 1433160.95 frames. ], batch size: 47, lr: 1.17e-02, grad_scale: 16.0 +2023-03-20 22:52:09,821 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 2.281e+02 2.630e+02 3.242e+02 8.955e+02, threshold=5.260e+02, percent-clipped=1.0 +2023-03-20 22:52:11,891 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37720.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:52:22,916 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 22:52:34,153 INFO [train.py:901] (1/2) Epoch 14, batch 1050, loss[loss=0.1737, simple_loss=0.2454, pruned_loss=0.05099, over 7322.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2422, pruned_loss=0.04785, over 1433385.82 frames. ], batch size: 61, lr: 1.17e-02, grad_scale: 16.0 +2023-03-20 22:52:46,477 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 22:52:47,174 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3673, 3.4066, 2.3676, 4.1853, 2.9865, 3.5164, 1.9888, 2.2036], + device='cuda:1'), covar=tensor([0.0290, 0.0557, 0.1963, 0.0402, 0.0385, 0.0887, 0.2846, 0.1969], + device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0230, 0.0303, 0.0225, 0.0242, 0.0232, 0.0270, 0.0289], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-20 22:52:51,545 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 22:52:57,597 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7968, 2.1562, 1.7627, 2.5743, 2.3398, 2.4088, 1.9333, 2.1478], + device='cuda:1'), covar=tensor([0.1914, 0.0770, 0.3016, 0.0524, 0.0076, 0.0075, 0.0116, 0.0174], + device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0213, 0.0261, 0.0245, 0.0123, 0.0120, 0.0144, 0.0167], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:53:00,402 INFO [train.py:901] (1/2) Epoch 14, batch 1100, loss[loss=0.1433, simple_loss=0.2176, pruned_loss=0.03447, over 7149.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2416, pruned_loss=0.04744, over 1434781.40 frames. ], batch size: 41, lr: 1.17e-02, grad_scale: 16.0 +2023-03-20 22:53:01,902 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.212e+02 2.615e+02 3.217e+02 6.533e+02, threshold=5.230e+02, percent-clipped=5.0 +2023-03-20 22:53:20,087 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 22:53:20,603 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 22:53:26,184 INFO [train.py:901] (1/2) Epoch 14, batch 1150, loss[loss=0.1679, simple_loss=0.2377, pruned_loss=0.04904, over 7316.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2408, pruned_loss=0.04689, over 1437391.99 frames. ], batch size: 75, lr: 1.17e-02, grad_scale: 16.0 +2023-03-20 22:53:32,739 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 22:53:33,245 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 22:53:35,508 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.35 vs. limit=5.0 +2023-03-20 22:53:41,569 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 +2023-03-20 22:53:51,815 INFO [train.py:901] (1/2) Epoch 14, batch 1200, loss[loss=0.1592, simple_loss=0.2444, pruned_loss=0.03697, over 7341.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2419, pruned_loss=0.04734, over 1437493.73 frames. ], batch size: 61, lr: 1.17e-02, grad_scale: 16.0 +2023-03-20 22:53:53,338 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.643e+02 2.300e+02 2.689e+02 3.229e+02 6.828e+02, threshold=5.378e+02, percent-clipped=3.0 +2023-03-20 22:53:57,412 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37924.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:54:05,068 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 22:54:11,385 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2700, 1.0889, 1.4638, 1.4778, 1.6226, 1.6496, 1.3241, 1.3621], + device='cuda:1'), covar=tensor([0.0781, 0.2713, 0.0713, 0.1078, 0.1672, 0.1973, 0.1395, 0.4685], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0046, 0.0035, 0.0031, 0.0038, 0.0036, 0.0044, 0.0039], + device='cuda:1'), out_proj_covar=tensor([1.0007e-04, 1.1285e-04, 8.9653e-05, 8.7168e-05, 9.9593e-05, 9.8123e-05, + 1.1122e-04, 1.0287e-04], device='cuda:1') +2023-03-20 22:54:18,270 INFO [train.py:901] (1/2) Epoch 14, batch 1250, loss[loss=0.1996, simple_loss=0.2671, pruned_loss=0.06602, over 7313.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2413, pruned_loss=0.04713, over 1438849.20 frames. ], batch size: 80, lr: 1.17e-02, grad_scale: 16.0 +2023-03-20 22:54:22,836 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=37972.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:54:26,501 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37979.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:54:30,368 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 22:54:34,431 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 22:54:35,412 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 22:54:39,124 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 +2023-03-20 22:54:43,645 INFO [train.py:901] (1/2) Epoch 14, batch 1300, loss[loss=0.1605, simple_loss=0.2251, pruned_loss=0.04793, over 7167.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2407, pruned_loss=0.0473, over 1438445.56 frames. ], batch size: 39, lr: 1.17e-02, grad_scale: 16.0 +2023-03-20 22:54:45,115 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.177e+02 2.872e+02 3.523e+02 8.377e+02, threshold=5.744e+02, percent-clipped=2.0 +2023-03-20 22:54:47,274 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38020.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:54:56,474 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0894, 1.1850, 1.0694, 1.2074, 1.1507, 1.1582, 0.9927, 0.8385], + device='cuda:1'), covar=tensor([0.0134, 0.0129, 0.0215, 0.0125, 0.0156, 0.0150, 0.0137, 0.0132], + device='cuda:1'), in_proj_covar=tensor([0.0023, 0.0021, 0.0021, 0.0020, 0.0023, 0.0021, 0.0022, 0.0028], + device='cuda:1'), out_proj_covar=tensor([2.7757e-05, 2.4370e-05, 2.5070e-05, 2.1926e-05, 2.8356e-05, 2.3991e-05, + 2.5160e-05, 3.3708e-05], device='cuda:1') +2023-03-20 22:54:58,003 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38040.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:54:58,787 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 22:55:01,295 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 22:55:05,239 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 22:55:09,670 INFO [train.py:901] (1/2) Epoch 14, batch 1350, loss[loss=0.1553, simple_loss=0.2396, pruned_loss=0.03554, over 7288.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2417, pruned_loss=0.04764, over 1441107.25 frames. ], batch size: 66, lr: 1.17e-02, grad_scale: 16.0 +2023-03-20 22:55:12,215 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=38068.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:55:15,570 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 22:55:23,809 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2255, 2.5735, 2.1154, 3.7514, 1.5878, 3.2646, 1.3800, 2.7771], + device='cuda:1'), covar=tensor([0.0066, 0.0689, 0.1514, 0.0046, 0.3286, 0.0076, 0.0971, 0.0165], + device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0261, 0.0298, 0.0150, 0.0283, 0.0159, 0.0266, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 22:55:27,746 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0730, 3.9552, 3.5565, 3.3795, 3.4141, 2.2925, 2.0795, 4.0643], + device='cuda:1'), covar=tensor([0.0025, 0.0039, 0.0061, 0.0057, 0.0077, 0.0375, 0.0407, 0.0033], + device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0065, 0.0085, 0.0073, 0.0089, 0.0111, 0.0111, 0.0076], + device='cuda:1'), out_proj_covar=tensor([9.5024e-05, 9.4697e-05, 1.1190e-04, 1.0293e-04, 1.1871e-04, 1.4949e-04, + 1.4877e-04, 1.0053e-04], device='cuda:1') +2023-03-20 22:55:35,273 INFO [train.py:901] (1/2) Epoch 14, batch 1400, loss[loss=0.1699, simple_loss=0.2404, pruned_loss=0.04972, over 7335.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2412, pruned_loss=0.04759, over 1440990.04 frames. ], batch size: 54, lr: 1.17e-02, grad_scale: 16.0 +2023-03-20 22:55:36,728 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.367e+02 2.853e+02 3.401e+02 7.956e+02, threshold=5.705e+02, percent-clipped=5.0 +2023-03-20 22:55:44,961 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.9622, 0.9514, 1.4791, 1.3780, 1.3372, 1.2675, 1.1711, 1.2607], + device='cuda:1'), covar=tensor([0.1644, 0.2632, 0.0642, 0.1900, 0.1247, 1.0454, 0.1321, 0.2724], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0047, 0.0035, 0.0032, 0.0039, 0.0037, 0.0044, 0.0040], + device='cuda:1'), out_proj_covar=tensor([1.0056e-04, 1.1534e-04, 9.0730e-05, 9.0102e-05, 1.0085e-04, 1.0164e-04, + 1.1320e-04, 1.0495e-04], device='cuda:1') +2023-03-20 22:55:49,493 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 22:56:00,688 INFO [train.py:901] (1/2) Epoch 14, batch 1450, loss[loss=0.1785, simple_loss=0.2561, pruned_loss=0.05051, over 7214.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2422, pruned_loss=0.04807, over 1440042.62 frames. ], batch size: 93, lr: 1.17e-02, grad_scale: 16.0 +2023-03-20 22:56:00,823 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9062, 3.2488, 3.8047, 3.8530, 3.9770, 3.7703, 3.7950, 3.5363], + device='cuda:1'), covar=tensor([0.0025, 0.0095, 0.0032, 0.0027, 0.0022, 0.0031, 0.0035, 0.0053], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0046, 0.0041, 0.0038, 0.0039, 0.0041, 0.0045, 0.0050], + device='cuda:1'), out_proj_covar=tensor([7.8699e-05, 1.2484e-04, 1.0845e-04, 9.0868e-05, 9.4839e-05, 1.0006e-04, + 1.1706e-04, 1.2600e-04], device='cuda:1') +2023-03-20 22:56:13,069 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 22:56:26,653 INFO [train.py:901] (1/2) Epoch 14, batch 1500, loss[loss=0.1691, simple_loss=0.2422, pruned_loss=0.04796, over 7311.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2423, pruned_loss=0.048, over 1439932.29 frames. ], batch size: 83, lr: 1.17e-02, grad_scale: 16.0 +2023-03-20 22:56:28,105 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.257e+02 2.452e+02 2.832e+02 3.488e+02 6.170e+02, threshold=5.665e+02, percent-clipped=2.0 +2023-03-20 22:56:30,097 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 22:56:36,997 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 +2023-03-20 22:56:52,230 INFO [train.py:901] (1/2) Epoch 14, batch 1550, loss[loss=0.1628, simple_loss=0.2385, pruned_loss=0.04356, over 7285.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2421, pruned_loss=0.04744, over 1442767.46 frames. ], batch size: 57, lr: 1.17e-02, grad_scale: 16.0 +2023-03-20 22:56:54,269 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 22:56:55,379 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3168, 2.7119, 2.1868, 2.5258, 2.7130, 2.0596, 2.4850, 2.4256], + device='cuda:1'), covar=tensor([0.1082, 0.1218, 0.1188, 0.1584, 0.1325, 0.0998, 0.1436, 0.1483], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0041, 0.0047, 0.0041, 0.0041, 0.0042, 0.0044, 0.0039], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-20 22:57:18,490 INFO [train.py:901] (1/2) Epoch 14, batch 1600, loss[loss=0.1702, simple_loss=0.2474, pruned_loss=0.04644, over 7269.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2417, pruned_loss=0.04725, over 1443023.57 frames. ], batch size: 52, lr: 1.16e-02, grad_scale: 16.0 +2023-03-20 22:57:19,972 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.415e+02 2.154e+02 2.623e+02 3.275e+02 5.102e+02, threshold=5.247e+02, percent-clipped=0.0 +2023-03-20 22:57:26,047 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 22:57:27,025 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 22:57:28,213 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6938, 3.3255, 3.2892, 3.4080, 2.8758, 2.6731, 3.6028, 2.8652], + device='cuda:1'), covar=tensor([0.0268, 0.0206, 0.0272, 0.0280, 0.0377, 0.0491, 0.0230, 0.0979], + device='cuda:1'), in_proj_covar=tensor([0.0292, 0.0296, 0.0246, 0.0311, 0.0301, 0.0296, 0.0298, 0.0294], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 22:57:29,597 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38335.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:57:30,055 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 22:57:30,640 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1262, 4.0049, 3.6824, 3.3480, 3.4273, 2.1950, 1.6645, 4.1037], + device='cuda:1'), covar=tensor([0.0045, 0.0057, 0.0083, 0.0083, 0.0117, 0.0537, 0.0665, 0.0062], + device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0066, 0.0086, 0.0074, 0.0091, 0.0111, 0.0113, 0.0076], + device='cuda:1'), out_proj_covar=tensor([9.5381e-05, 9.5519e-05, 1.1339e-04, 1.0358e-04, 1.2009e-04, 1.4918e-04, + 1.5120e-04, 1.0022e-04], device='cuda:1') +2023-03-20 22:57:40,021 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 22:57:42,167 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38360.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:57:43,450 INFO [train.py:901] (1/2) Epoch 14, batch 1650, loss[loss=0.1787, simple_loss=0.2472, pruned_loss=0.05507, over 7343.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2424, pruned_loss=0.04764, over 1444265.93 frames. ], batch size: 63, lr: 1.16e-02, grad_scale: 8.0 +2023-03-20 22:57:44,445 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 22:57:52,027 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 22:58:09,232 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4311, 3.2588, 3.0197, 3.1594, 2.5312, 2.4061, 3.4946, 2.6445], + device='cuda:1'), covar=tensor([0.0300, 0.0260, 0.0295, 0.0301, 0.0384, 0.0517, 0.0330, 0.0812], + device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0295, 0.0246, 0.0310, 0.0298, 0.0293, 0.0299, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 22:58:09,560 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 22:58:09,994 INFO [train.py:901] (1/2) Epoch 14, batch 1700, loss[loss=0.2184, simple_loss=0.2801, pruned_loss=0.07832, over 6680.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.242, pruned_loss=0.04755, over 1444757.95 frames. ], batch size: 107, lr: 1.16e-02, grad_scale: 8.0 +2023-03-20 22:58:12,009 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.225e+02 2.418e+02 2.831e+02 3.650e+02 6.823e+02, threshold=5.661e+02, percent-clipped=1.0 +2023-03-20 22:58:13,077 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 22:58:14,206 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38421.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 22:58:21,277 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.9259, 4.0200, 4.5732, 4.4155, 4.4201, 4.4634, 4.6547, 4.3459], + device='cuda:1'), covar=tensor([0.0056, 0.0142, 0.0081, 0.0100, 0.0249, 0.0081, 0.0102, 0.0103], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0069, 0.0070, 0.0060, 0.0120, 0.0078, 0.0073, 0.0074], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:58:23,759 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 22:58:35,222 INFO [train.py:901] (1/2) Epoch 14, batch 1750, loss[loss=0.1718, simple_loss=0.2493, pruned_loss=0.04717, over 7290.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2422, pruned_loss=0.04756, over 1443526.92 frames. ], batch size: 68, lr: 1.16e-02, grad_scale: 8.0 +2023-03-20 22:58:48,463 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 22:58:48,992 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 22:58:55,481 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-20 22:59:00,786 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8466, 2.4446, 1.8419, 2.9412, 2.1674, 2.1816, 2.3372, 1.9351], + device='cuda:1'), covar=tensor([0.1565, 0.0611, 0.2474, 0.0390, 0.0058, 0.0048, 0.0160, 0.0195], + device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0222, 0.0268, 0.0254, 0.0129, 0.0124, 0.0152, 0.0177], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:59:01,588 INFO [train.py:901] (1/2) Epoch 14, batch 1800, loss[loss=0.2098, simple_loss=0.2734, pruned_loss=0.07313, over 6593.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2421, pruned_loss=0.04732, over 1442491.68 frames. ], batch size: 106, lr: 1.16e-02, grad_scale: 8.0 +2023-03-20 22:59:03,607 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.490e+02 2.294e+02 2.677e+02 3.178e+02 6.312e+02, threshold=5.355e+02, percent-clipped=2.0 +2023-03-20 22:59:11,143 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 22:59:11,827 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7336, 2.5141, 2.4108, 3.6323, 1.5754, 3.6697, 1.3887, 3.0588], + device='cuda:1'), covar=tensor([0.0073, 0.0836, 0.1491, 0.0053, 0.3781, 0.0105, 0.0964, 0.0254], + device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0260, 0.0296, 0.0150, 0.0281, 0.0159, 0.0263, 0.0214], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 22:59:25,438 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 22:59:27,464 INFO [train.py:901] (1/2) Epoch 14, batch 1850, loss[loss=0.1504, simple_loss=0.2296, pruned_loss=0.03564, over 7318.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2413, pruned_loss=0.04666, over 1443700.74 frames. ], batch size: 44, lr: 1.16e-02, grad_scale: 8.0 +2023-03-20 22:59:35,641 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 22:59:43,873 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9582, 2.5031, 1.7123, 2.7891, 2.4114, 2.4214, 2.2762, 1.8041], + device='cuda:1'), covar=tensor([0.1855, 0.0800, 0.3392, 0.0513, 0.0074, 0.0054, 0.0152, 0.0223], + device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0226, 0.0274, 0.0260, 0.0132, 0.0126, 0.0154, 0.0178], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 22:59:52,573 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 22:59:53,599 INFO [train.py:901] (1/2) Epoch 14, batch 1900, loss[loss=0.1577, simple_loss=0.2359, pruned_loss=0.03971, over 7338.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2412, pruned_loss=0.0466, over 1441636.24 frames. ], batch size: 54, lr: 1.16e-02, grad_scale: 8.0 +2023-03-20 22:59:55,554 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.311e+02 2.784e+02 3.401e+02 7.573e+02, threshold=5.568e+02, percent-clipped=6.0 +2023-03-20 23:00:03,743 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8603, 2.3462, 2.9668, 2.8748, 2.9219, 2.6572, 2.3964, 2.9000], + device='cuda:1'), covar=tensor([0.1530, 0.0436, 0.1165, 0.1177, 0.0969, 0.1459, 0.2542, 0.1522], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0041, 0.0036, 0.0037, 0.0035, 0.0034, 0.0049, 0.0037], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:00:04,718 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38635.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:00:18,440 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 23:00:19,859 INFO [train.py:901] (1/2) Epoch 14, batch 1950, loss[loss=0.1847, simple_loss=0.2543, pruned_loss=0.05758, over 7276.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2407, pruned_loss=0.04706, over 1439562.04 frames. ], batch size: 64, lr: 1.16e-02, grad_scale: 8.0 +2023-03-20 23:00:28,937 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 23:00:29,959 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=38683.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:00:32,946 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2704, 2.3410, 2.1641, 3.7680, 1.5068, 3.2645, 1.3767, 2.9603], + device='cuda:1'), covar=tensor([0.0083, 0.0975, 0.1572, 0.0055, 0.4059, 0.0110, 0.1056, 0.0246], + device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0259, 0.0291, 0.0151, 0.0281, 0.0160, 0.0263, 0.0213], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 23:00:33,784 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 23:00:34,303 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 23:00:36,534 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7717, 2.6671, 2.5096, 2.6439, 2.1844, 2.0450, 2.7667, 2.1145], + device='cuda:1'), covar=tensor([0.0226, 0.0262, 0.0247, 0.0307, 0.0358, 0.0485, 0.0282, 0.0938], + device='cuda:1'), in_proj_covar=tensor([0.0287, 0.0290, 0.0243, 0.0306, 0.0294, 0.0293, 0.0297, 0.0291], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 23:00:44,915 INFO [train.py:901] (1/2) Epoch 14, batch 2000, loss[loss=0.1366, simple_loss=0.2132, pruned_loss=0.02999, over 7340.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2403, pruned_loss=0.04694, over 1439294.92 frames. ], batch size: 44, lr: 1.16e-02, grad_scale: 8.0 +2023-03-20 23:00:46,497 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38716.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:00:46,882 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.473e+02 2.269e+02 2.724e+02 3.233e+02 7.728e+02, threshold=5.449e+02, percent-clipped=4.0 +2023-03-20 23:00:47,545 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5070, 3.1596, 3.6112, 3.3083, 3.2297, 3.3694, 2.9694, 3.7591], + device='cuda:1'), covar=tensor([0.1132, 0.0642, 0.1091, 0.1561, 0.1051, 0.1021, 0.2573, 0.0763], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0041, 0.0036, 0.0036, 0.0034, 0.0033, 0.0048, 0.0036], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:00:48,869 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 +2023-03-20 23:00:50,468 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 23:01:02,775 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 23:01:10,339 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 23:01:11,303 INFO [train.py:901] (1/2) Epoch 14, batch 2050, loss[loss=0.1583, simple_loss=0.2259, pruned_loss=0.04533, over 7310.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2395, pruned_loss=0.04655, over 1438410.21 frames. ], batch size: 49, lr: 1.16e-02, grad_scale: 8.0 +2023-03-20 23:01:36,579 INFO [train.py:901] (1/2) Epoch 14, batch 2100, loss[loss=0.1876, simple_loss=0.2619, pruned_loss=0.05663, over 7296.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2407, pruned_loss=0.04704, over 1436718.43 frames. ], batch size: 86, lr: 1.16e-02, grad_scale: 8.0 +2023-03-20 23:01:39,202 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.353e+02 2.685e+02 3.230e+02 6.902e+02, threshold=5.371e+02, percent-clipped=1.0 +2023-03-20 23:01:42,252 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38823.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:01:42,641 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 23:01:46,014 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.01 vs. limit=5.0 +2023-03-20 23:01:46,186 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 23:01:53,167 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9368, 3.4753, 3.7267, 3.7756, 3.8009, 3.9085, 3.8649, 3.6799], + device='cuda:1'), covar=tensor([0.0021, 0.0074, 0.0035, 0.0039, 0.0026, 0.0026, 0.0034, 0.0050], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0046, 0.0041, 0.0038, 0.0040, 0.0040, 0.0044, 0.0051], + device='cuda:1'), out_proj_covar=tensor([7.8380e-05, 1.2323e-04, 1.0879e-04, 9.0369e-05, 9.6843e-05, 9.5575e-05, + 1.1516e-04, 1.2598e-04], device='cuda:1') +2023-03-20 23:02:02,525 INFO [train.py:901] (1/2) Epoch 14, batch 2150, loss[loss=0.1535, simple_loss=0.2351, pruned_loss=0.03593, over 7324.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2415, pruned_loss=0.04709, over 1439862.34 frames. ], batch size: 49, lr: 1.16e-02, grad_scale: 8.0 +2023-03-20 23:02:13,059 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38884.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:02:16,016 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5279, 3.7779, 3.5712, 3.6395, 3.4126, 3.6975, 4.0114, 4.0278], + device='cuda:1'), covar=tensor([0.0231, 0.0179, 0.0215, 0.0265, 0.0373, 0.0257, 0.0218, 0.0156], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0102, 0.0095, 0.0104, 0.0097, 0.0085, 0.0081, 0.0080], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:02:21,326 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1087, 3.0714, 2.1567, 3.8410, 2.4250, 3.0599, 1.8338, 2.0286], + device='cuda:1'), covar=tensor([0.0208, 0.0611, 0.1667, 0.0366, 0.0256, 0.0238, 0.2103, 0.1401], + device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0229, 0.0299, 0.0228, 0.0247, 0.0232, 0.0269, 0.0285], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-20 23:02:28,847 INFO [train.py:901] (1/2) Epoch 14, batch 2200, loss[loss=0.1883, simple_loss=0.2608, pruned_loss=0.05788, over 7337.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2416, pruned_loss=0.04734, over 1440427.44 frames. ], batch size: 54, lr: 1.16e-02, grad_scale: 8.0 +2023-03-20 23:02:30,375 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 23:02:30,841 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.439e+02 2.207e+02 2.666e+02 3.268e+02 4.931e+02, threshold=5.333e+02, percent-clipped=0.0 +2023-03-20 23:02:44,953 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8277, 2.4003, 1.9129, 2.5612, 2.2547, 2.2629, 2.2328, 2.1109], + device='cuda:1'), covar=tensor([0.1814, 0.0679, 0.2770, 0.0294, 0.0064, 0.0049, 0.0130, 0.0164], + device='cuda:1'), in_proj_covar=tensor([0.0293, 0.0229, 0.0280, 0.0263, 0.0136, 0.0132, 0.0158, 0.0183], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:02:53,828 INFO [train.py:901] (1/2) Epoch 14, batch 2250, loss[loss=0.1231, simple_loss=0.1912, pruned_loss=0.0275, over 7014.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2414, pruned_loss=0.04723, over 1441111.15 frames. ], batch size: 35, lr: 1.15e-02, grad_scale: 8.0 +2023-03-20 23:03:04,532 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 23:03:05,023 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 23:03:17,858 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 23:03:20,293 INFO [train.py:901] (1/2) Epoch 14, batch 2300, loss[loss=0.1563, simple_loss=0.238, pruned_loss=0.03727, over 7362.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.241, pruned_loss=0.04694, over 1441021.25 frames. ], batch size: 73, lr: 1.15e-02, grad_scale: 8.0 +2023-03-20 23:03:21,968 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39016.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:03:22,306 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.345e+02 2.210e+02 2.631e+02 3.311e+02 5.841e+02, threshold=5.262e+02, percent-clipped=4.0 +2023-03-20 23:03:25,966 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4732, 4.8513, 4.8214, 4.7956, 4.7133, 4.6531, 4.9344, 4.5934], + device='cuda:1'), covar=tensor([0.0903, 0.1118, 0.0909, 0.1149, 0.0814, 0.0728, 0.0936, 0.1123], + device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0197, 0.0137, 0.0145, 0.0122, 0.0177, 0.0154, 0.0120], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:03:45,471 INFO [train.py:901] (1/2) Epoch 14, batch 2350, loss[loss=0.1619, simple_loss=0.2464, pruned_loss=0.03876, over 7223.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2404, pruned_loss=0.0465, over 1442296.93 frames. ], batch size: 93, lr: 1.15e-02, grad_scale: 8.0 +2023-03-20 23:03:46,050 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=39064.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:04:01,197 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6196, 5.1464, 5.2042, 5.1611, 4.9898, 4.7292, 5.2176, 4.9798], + device='cuda:1'), covar=tensor([0.0408, 0.0377, 0.0349, 0.0372, 0.0294, 0.0308, 0.0352, 0.0472], + device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0201, 0.0141, 0.0146, 0.0123, 0.0180, 0.0157, 0.0121], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:04:06,173 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 23:04:09,269 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9422, 3.7901, 3.4579, 3.2673, 3.2135, 2.1112, 1.7174, 3.9216], + device='cuda:1'), covar=tensor([0.0027, 0.0050, 0.0057, 0.0059, 0.0079, 0.0388, 0.0495, 0.0036], + device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0068, 0.0088, 0.0074, 0.0093, 0.0112, 0.0115, 0.0077], + device='cuda:1'), out_proj_covar=tensor([9.6394e-05, 9.8711e-05, 1.1591e-04, 1.0394e-04, 1.2260e-04, 1.5094e-04, + 1.5431e-04, 1.0053e-04], device='cuda:1') +2023-03-20 23:04:11,733 INFO [train.py:901] (1/2) Epoch 14, batch 2400, loss[loss=0.1398, simple_loss=0.1986, pruned_loss=0.04045, over 6209.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2412, pruned_loss=0.04696, over 1443900.69 frames. ], batch size: 27, lr: 1.15e-02, grad_scale: 8.0 +2023-03-20 23:04:12,259 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 23:04:13,735 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.544e+02 2.325e+02 2.690e+02 3.201e+02 5.267e+02, threshold=5.381e+02, percent-clipped=1.0 +2023-03-20 23:04:22,285 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 23:04:22,640 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 +2023-03-20 23:04:24,751 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 23:04:37,375 INFO [train.py:901] (1/2) Epoch 14, batch 2450, loss[loss=0.1417, simple_loss=0.2146, pruned_loss=0.03442, over 6970.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2411, pruned_loss=0.04695, over 1445433.60 frames. ], batch size: 35, lr: 1.15e-02, grad_scale: 8.0 +2023-03-20 23:04:46,217 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39179.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:04:49,252 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39185.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:04:52,559 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 23:05:03,366 INFO [train.py:901] (1/2) Epoch 14, batch 2500, loss[loss=0.1703, simple_loss=0.24, pruned_loss=0.05035, over 7326.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2404, pruned_loss=0.04678, over 1444188.08 frames. ], batch size: 75, lr: 1.15e-02, grad_scale: 8.0 +2023-03-20 23:05:05,414 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.736e+02 2.502e+02 3.022e+02 3.546e+02 7.037e+02, threshold=6.044e+02, percent-clipped=3.0 +2023-03-20 23:05:06,305 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-03-20 23:05:18,649 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 23:05:20,290 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39246.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:05:29,213 INFO [train.py:901] (1/2) Epoch 14, batch 2550, loss[loss=0.1494, simple_loss=0.2277, pruned_loss=0.03551, over 7239.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2399, pruned_loss=0.04671, over 1441168.29 frames. ], batch size: 47, lr: 1.15e-02, grad_scale: 8.0 +2023-03-20 23:05:40,512 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39284.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:05:54,845 INFO [train.py:901] (1/2) Epoch 14, batch 2600, loss[loss=0.1552, simple_loss=0.2233, pruned_loss=0.04361, over 7132.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2386, pruned_loss=0.04602, over 1441184.12 frames. ], batch size: 41, lr: 1.15e-02, grad_scale: 8.0 +2023-03-20 23:05:56,767 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.181e+02 2.670e+02 3.207e+02 6.854e+02, threshold=5.339e+02, percent-clipped=3.0 +2023-03-20 23:06:03,881 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-20 23:06:07,043 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-20 23:06:09,179 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8800, 2.3832, 2.8236, 2.7884, 2.7840, 2.7083, 2.2196, 2.6752], + device='cuda:1'), covar=tensor([0.0967, 0.0684, 0.1188, 0.1810, 0.1076, 0.1188, 0.2448, 0.2208], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0044, 0.0039, 0.0040, 0.0036, 0.0034, 0.0052, 0.0040], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:06:10,660 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39345.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:06:19,253 INFO [train.py:901] (1/2) Epoch 14, batch 2650, loss[loss=0.1373, simple_loss=0.2017, pruned_loss=0.03639, over 7023.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.24, pruned_loss=0.0468, over 1441474.03 frames. ], batch size: 35, lr: 1.15e-02, grad_scale: 8.0 +2023-03-20 23:06:44,700 INFO [train.py:901] (1/2) Epoch 14, batch 2700, loss[loss=0.1796, simple_loss=0.2462, pruned_loss=0.05652, over 7286.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2403, pruned_loss=0.04677, over 1443114.59 frames. ], batch size: 70, lr: 1.15e-02, grad_scale: 8.0 +2023-03-20 23:06:46,682 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.412e+02 2.100e+02 2.474e+02 3.007e+02 6.884e+02, threshold=4.948e+02, percent-clipped=2.0 +2023-03-20 23:06:48,686 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6156, 3.3145, 3.3814, 3.2687, 3.2234, 3.2014, 3.5364, 3.1453], + device='cuda:1'), covar=tensor([0.0115, 0.0204, 0.0118, 0.0186, 0.0392, 0.0114, 0.0159, 0.0176], + device='cuda:1'), in_proj_covar=tensor([0.0064, 0.0072, 0.0072, 0.0062, 0.0124, 0.0081, 0.0077, 0.0077], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:06:50,695 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3315, 2.2071, 1.9932, 1.9665, 1.4800, 1.7725, 1.7382, 1.5379], + device='cuda:1'), covar=tensor([0.0463, 0.0324, 0.0145, 0.0124, 0.0983, 0.0285, 0.0241, 0.0276], + device='cuda:1'), in_proj_covar=tensor([0.0022, 0.0023, 0.0022, 0.0021, 0.0021, 0.0021, 0.0023, 0.0024], + device='cuda:1'), out_proj_covar=tensor([5.5454e-05, 5.6677e-05, 5.2041e-05, 4.8400e-05, 5.3448e-05, 5.2388e-05, + 5.5534e-05, 6.0192e-05], device='cuda:1') +2023-03-20 23:06:56,714 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-03-20 23:07:09,771 INFO [train.py:901] (1/2) Epoch 14, batch 2750, loss[loss=0.1378, simple_loss=0.212, pruned_loss=0.0318, over 7347.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2411, pruned_loss=0.04745, over 1442231.32 frames. ], batch size: 44, lr: 1.15e-02, grad_scale: 8.0 +2023-03-20 23:07:09,919 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39463.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:07:17,626 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39479.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:07:33,827 INFO [train.py:901] (1/2) Epoch 14, batch 2800, loss[loss=0.1575, simple_loss=0.2318, pruned_loss=0.04156, over 7321.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2417, pruned_loss=0.04753, over 1442522.00 frames. ], batch size: 49, lr: 1.15e-02, grad_scale: 8.0 +2023-03-20 23:07:35,753 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.700e+02 2.191e+02 2.661e+02 3.280e+02 1.218e+03, threshold=5.323e+02, percent-clipped=4.0 +2023-03-20 23:07:39,317 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39524.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:07:40,721 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=39527.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:07:42,745 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1869, 4.4674, 4.2665, 4.3944, 4.2090, 4.3682, 4.7023, 4.8130], + device='cuda:1'), covar=tensor([0.0212, 0.0153, 0.0163, 0.0143, 0.0311, 0.0162, 0.0225, 0.0156], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0103, 0.0093, 0.0106, 0.0097, 0.0085, 0.0079, 0.0080], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:07:59,847 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 23:07:59,900 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 23:07:59,926 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 23:07:59,944 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 23:08:00,113 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 23:08:00,128 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 23:08:00,138 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 23:08:00,234 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 23:08:00,306 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 23:08:00,321 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 23:08:00,336 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 23:08:00,827 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 23:08:00,833 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 23:08:00,840 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 23:08:00,976 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 23:08:00,978 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 23:08:01,036 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 23:08:09,210 INFO [train.py:901] (1/2) Epoch 15, batch 0, loss[loss=0.1447, simple_loss=0.2232, pruned_loss=0.03313, over 7339.00 frames. ], tot_loss[loss=0.1447, simple_loss=0.2232, pruned_loss=0.03313, over 7339.00 frames. ], batch size: 54, lr: 1.11e-02, grad_scale: 8.0 +2023-03-20 23:08:09,210 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 23:08:13,261 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5062, 2.7690, 2.3908, 2.6150, 2.7338, 2.2877, 2.7495, 2.6800], + device='cuda:1'), covar=tensor([0.0902, 0.0664, 0.1106, 0.1215, 0.1020, 0.1108, 0.1516, 0.1295], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0041, 0.0047, 0.0042, 0.0043, 0.0043, 0.0043, 0.0040], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:08:29,216 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9763, 3.5176, 3.5442, 3.6831, 3.6786, 3.5164, 3.7019, 3.4516], + device='cuda:1'), covar=tensor([0.0098, 0.0166, 0.0162, 0.0148, 0.0370, 0.0130, 0.0191, 0.0151], + device='cuda:1'), in_proj_covar=tensor([0.0063, 0.0070, 0.0071, 0.0061, 0.0122, 0.0079, 0.0075, 0.0075], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:08:31,435 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4191, 1.8762, 1.7022, 1.6586, 1.4503, 1.6119, 1.9217, 1.5277], + device='cuda:1'), covar=tensor([0.0227, 0.0399, 0.0219, 0.0127, 0.0607, 0.0384, 0.0161, 0.0287], + device='cuda:1'), in_proj_covar=tensor([0.0022, 0.0023, 0.0022, 0.0021, 0.0021, 0.0021, 0.0023, 0.0024], + device='cuda:1'), out_proj_covar=tensor([5.5820e-05, 5.7003e-05, 5.1662e-05, 4.8850e-05, 5.3773e-05, 5.2706e-05, + 5.4645e-05, 6.0357e-05], device='cuda:1') +2023-03-20 23:08:33,616 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7868, 3.6898, 2.8594, 3.2875, 3.2408, 2.1714, 1.6851, 3.8239], + device='cuda:1'), covar=tensor([0.0019, 0.0030, 0.0083, 0.0058, 0.0057, 0.0407, 0.0481, 0.0027], + device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0067, 0.0087, 0.0073, 0.0093, 0.0111, 0.0114, 0.0077], + device='cuda:1'), out_proj_covar=tensor([9.3879e-05, 9.7424e-05, 1.1427e-04, 1.0210e-04, 1.2239e-04, 1.4984e-04, + 1.5256e-04, 1.0034e-04], device='cuda:1') +2023-03-20 23:08:35,274 INFO [train.py:935] (1/2) Epoch 15, validation: loss=0.1689, simple_loss=0.2557, pruned_loss=0.04109, over 1622729.00 frames. +2023-03-20 23:08:35,274 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-20 23:08:37,359 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39541.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:08:42,283 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 23:08:52,831 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 23:08:57,976 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2307, 2.1538, 2.0434, 3.4760, 1.4567, 3.2091, 1.2534, 3.0010], + device='cuda:1'), covar=tensor([0.0069, 0.0825, 0.1733, 0.0056, 0.4130, 0.0089, 0.1188, 0.0205], + device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0259, 0.0302, 0.0152, 0.0282, 0.0160, 0.0267, 0.0213], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 23:08:59,293 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 23:09:00,284 INFO [train.py:901] (1/2) Epoch 15, batch 50, loss[loss=0.1549, simple_loss=0.2312, pruned_loss=0.0393, over 7200.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2422, pruned_loss=0.04703, over 327238.91 frames. ], batch size: 45, lr: 1.11e-02, grad_scale: 8.0 +2023-03-20 23:09:01,284 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 23:09:04,286 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 23:09:16,182 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.268e+02 2.158e+02 2.629e+02 3.348e+02 6.220e+02, threshold=5.258e+02, percent-clipped=5.0 +2023-03-20 23:09:21,413 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 23:09:21,900 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 23:09:26,886 INFO [train.py:901] (1/2) Epoch 15, batch 100, loss[loss=0.1866, simple_loss=0.2482, pruned_loss=0.06244, over 7307.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2418, pruned_loss=0.04653, over 574720.40 frames. ], batch size: 83, lr: 1.11e-02, grad_scale: 8.0 +2023-03-20 23:09:27,991 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39639.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 23:09:28,454 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39640.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:09:35,012 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3003, 2.2279, 2.1144, 3.5854, 1.5181, 3.2437, 1.3758, 3.0469], + device='cuda:1'), covar=tensor([0.0065, 0.0915, 0.1849, 0.0067, 0.4295, 0.0085, 0.0993, 0.0196], + device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0262, 0.0305, 0.0153, 0.0285, 0.0161, 0.0270, 0.0215], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 23:09:46,001 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9141, 3.7455, 3.2760, 3.2603, 3.0412, 2.0717, 1.4249, 3.9204], + device='cuda:1'), covar=tensor([0.0048, 0.0073, 0.0119, 0.0085, 0.0149, 0.0547, 0.0695, 0.0063], + device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0067, 0.0088, 0.0073, 0.0093, 0.0113, 0.0116, 0.0077], + device='cuda:1'), out_proj_covar=tensor([9.5218e-05, 9.7815e-05, 1.1553e-04, 1.0241e-04, 1.2266e-04, 1.5116e-04, + 1.5472e-04, 1.0085e-04], device='cuda:1') +2023-03-20 23:09:47,590 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1252, 2.4870, 1.8586, 2.8933, 2.6588, 2.6793, 2.6029, 2.6519], + device='cuda:1'), covar=tensor([0.1752, 0.0668, 0.3358, 0.0539, 0.0091, 0.0063, 0.0166, 0.0263], + device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0226, 0.0272, 0.0258, 0.0135, 0.0128, 0.0154, 0.0180], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:09:52,621 INFO [train.py:901] (1/2) Epoch 15, batch 150, loss[loss=0.1736, simple_loss=0.2502, pruned_loss=0.04854, over 7297.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2407, pruned_loss=0.04614, over 767612.76 frames. ], batch size: 80, lr: 1.11e-02, grad_scale: 8.0 +2023-03-20 23:09:59,323 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39700.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 23:10:08,386 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.211e+02 2.562e+02 3.102e+02 7.128e+02, threshold=5.124e+02, percent-clipped=3.0 +2023-03-20 23:10:18,268 INFO [train.py:901] (1/2) Epoch 15, batch 200, loss[loss=0.1781, simple_loss=0.2539, pruned_loss=0.05118, over 7215.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2407, pruned_loss=0.04641, over 916041.84 frames. ], batch size: 93, lr: 1.11e-02, grad_scale: 8.0 +2023-03-20 23:10:22,287 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 23:10:26,812 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 23:10:32,825 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 23:10:33,424 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39767.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:10:43,919 INFO [train.py:901] (1/2) Epoch 15, batch 250, loss[loss=0.1614, simple_loss=0.2339, pruned_loss=0.04449, over 7278.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2396, pruned_loss=0.04591, over 1033529.42 frames. ], batch size: 57, lr: 1.11e-02, grad_scale: 8.0 +2023-03-20 23:10:46,550 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 23:10:58,985 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39815.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:10:59,823 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.607e+02 2.227e+02 2.655e+02 3.135e+02 9.236e+02, threshold=5.310e+02, percent-clipped=3.0 +2023-03-20 23:11:00,007 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2309, 2.7037, 2.3434, 2.6120, 2.6581, 2.1566, 2.7253, 2.5113], + device='cuda:1'), covar=tensor([0.1363, 0.0318, 0.0940, 0.0666, 0.0609, 0.0680, 0.0860, 0.0750], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0041, 0.0048, 0.0041, 0.0042, 0.0044, 0.0044, 0.0041], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:11:00,932 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39819.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:11:05,488 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39828.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:11:07,845 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 23:11:09,809 INFO [train.py:901] (1/2) Epoch 15, batch 300, loss[loss=0.198, simple_loss=0.2721, pruned_loss=0.06191, over 7119.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2394, pruned_loss=0.04599, over 1124899.45 frames. ], batch size: 98, lr: 1.11e-02, grad_scale: 8.0 +2023-03-20 23:11:11,854 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39841.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:11:12,407 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0476, 1.3259, 1.1869, 1.2095, 1.2328, 1.1809, 1.0851, 0.8527], + device='cuda:1'), covar=tensor([0.0157, 0.0072, 0.0129, 0.0101, 0.0118, 0.0076, 0.0115, 0.0115], + device='cuda:1'), in_proj_covar=tensor([0.0022, 0.0021, 0.0019, 0.0020, 0.0022, 0.0020, 0.0022, 0.0027], + device='cuda:1'), out_proj_covar=tensor([2.5640e-05, 2.3525e-05, 2.3205e-05, 2.2993e-05, 2.6334e-05, 2.2764e-05, + 2.4938e-05, 3.2670e-05], device='cuda:1') +2023-03-20 23:11:16,253 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 23:11:29,835 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 +2023-03-20 23:11:30,552 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39876.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:11:35,977 INFO [train.py:901] (1/2) Epoch 15, batch 350, loss[loss=0.1284, simple_loss=0.1886, pruned_loss=0.0341, over 6035.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.239, pruned_loss=0.04583, over 1194479.58 frames. ], batch size: 26, lr: 1.11e-02, grad_scale: 8.0 +2023-03-20 23:11:37,055 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=39889.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:11:50,883 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.396e+02 2.229e+02 2.631e+02 3.067e+02 9.607e+02, threshold=5.263e+02, percent-clipped=2.0 +2023-03-20 23:11:50,901 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 23:12:00,946 INFO [train.py:901] (1/2) Epoch 15, batch 400, loss[loss=0.1535, simple_loss=0.234, pruned_loss=0.0365, over 7293.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2392, pruned_loss=0.04626, over 1249358.54 frames. ], batch size: 86, lr: 1.11e-02, grad_scale: 8.0 +2023-03-20 23:12:02,503 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39940.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:12:26,852 INFO [train.py:901] (1/2) Epoch 15, batch 450, loss[loss=0.1519, simple_loss=0.2302, pruned_loss=0.03681, over 7362.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2396, pruned_loss=0.04644, over 1290756.69 frames. ], batch size: 73, lr: 1.11e-02, grad_scale: 8.0 +2023-03-20 23:12:27,416 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=39988.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:12:27,926 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2641, 4.8322, 4.9636, 4.8535, 4.8040, 4.3845, 4.9609, 4.7866], + device='cuda:1'), covar=tensor([0.0457, 0.0437, 0.0412, 0.0441, 0.0304, 0.0357, 0.0355, 0.0527], + device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0198, 0.0143, 0.0144, 0.0122, 0.0183, 0.0155, 0.0120], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:12:30,298 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 23:12:30,794 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 23:12:30,856 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39995.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 23:12:45,601 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.293e+02 2.173e+02 2.880e+02 3.214e+02 6.134e+02, threshold=5.760e+02, percent-clipped=1.0 +2023-03-20 23:12:53,209 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7027, 2.6374, 2.8638, 2.8150, 2.9401, 2.8644, 2.3933, 2.6962], + device='cuda:1'), covar=tensor([0.2359, 0.0457, 0.1807, 0.2182, 0.0989, 0.0950, 0.2475, 0.2097], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0043, 0.0038, 0.0040, 0.0036, 0.0034, 0.0052, 0.0040], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:12:56,077 INFO [train.py:901] (1/2) Epoch 15, batch 500, loss[loss=0.1564, simple_loss=0.2353, pruned_loss=0.0388, over 7355.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2398, pruned_loss=0.04625, over 1324350.11 frames. ], batch size: 73, lr: 1.10e-02, grad_scale: 8.0 +2023-03-20 23:13:06,675 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 23:13:08,631 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 23:13:09,575 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 23:13:11,547 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 23:13:16,607 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 23:13:16,721 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5050, 4.1770, 4.1234, 3.6225, 3.9807, 2.6901, 2.0941, 4.4439], + device='cuda:1'), covar=tensor([0.0020, 0.0067, 0.0041, 0.0059, 0.0048, 0.0327, 0.0442, 0.0029], + device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0066, 0.0085, 0.0071, 0.0091, 0.0109, 0.0112, 0.0075], + device='cuda:1'), out_proj_covar=tensor([9.4677e-05, 9.5680e-05, 1.1163e-04, 9.9604e-05, 1.1973e-04, 1.4488e-04, + 1.4926e-04, 9.7530e-05], device='cuda:1') +2023-03-20 23:13:21,552 INFO [train.py:901] (1/2) Epoch 15, batch 550, loss[loss=0.1797, simple_loss=0.2525, pruned_loss=0.05341, over 7308.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2387, pruned_loss=0.04575, over 1349593.68 frames. ], batch size: 83, lr: 1.10e-02, grad_scale: 8.0 +2023-03-20 23:13:27,424 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-20 23:13:28,015 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 23:13:36,021 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 23:13:36,485 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.123e+02 2.573e+02 3.365e+02 8.835e+02, threshold=5.146e+02, percent-clipped=2.0 +2023-03-20 23:13:38,249 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40119.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:13:40,140 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 23:13:40,221 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40123.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:13:43,331 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7610, 3.2517, 3.6503, 3.6543, 3.6167, 3.8398, 3.5773, 3.4901], + device='cuda:1'), covar=tensor([0.0022, 0.0074, 0.0031, 0.0029, 0.0031, 0.0022, 0.0040, 0.0055], + device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0045, 0.0041, 0.0037, 0.0039, 0.0039, 0.0042, 0.0049], + device='cuda:1'), out_proj_covar=tensor([7.4554e-05, 1.1859e-04, 1.0528e-04, 8.7248e-05, 9.1355e-05, 9.3312e-05, + 1.0911e-04, 1.1831e-04], device='cuda:1') +2023-03-20 23:13:47,346 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 23:13:47,832 INFO [train.py:901] (1/2) Epoch 15, batch 600, loss[loss=0.1365, simple_loss=0.2193, pruned_loss=0.02685, over 7113.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2384, pruned_loss=0.04527, over 1370564.16 frames. ], batch size: 41, lr: 1.10e-02, grad_scale: 8.0 +2023-03-20 23:14:02,882 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 23:14:02,938 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=40167.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:14:04,973 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40171.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:14:06,536 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2655, 1.6074, 1.4236, 1.3770, 1.6264, 1.1909, 1.1046, 1.1761], + device='cuda:1'), covar=tensor([0.0134, 0.0163, 0.0221, 0.0136, 0.0119, 0.0249, 0.0153, 0.0123], + device='cuda:1'), in_proj_covar=tensor([0.0022, 0.0022, 0.0020, 0.0021, 0.0023, 0.0021, 0.0022, 0.0028], + device='cuda:1'), out_proj_covar=tensor([2.6489e-05, 2.5106e-05, 2.4522e-05, 2.3990e-05, 2.7906e-05, 2.3825e-05, + 2.5432e-05, 3.3277e-05], device='cuda:1') +2023-03-20 23:14:11,946 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 23:14:12,941 INFO [train.py:901] (1/2) Epoch 15, batch 650, loss[loss=0.1637, simple_loss=0.2456, pruned_loss=0.04089, over 7352.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2382, pruned_loss=0.04486, over 1384702.85 frames. ], batch size: 63, lr: 1.10e-02, grad_scale: 8.0 +2023-03-20 23:14:13,814 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 +2023-03-20 23:14:17,022 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40195.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:14:20,888 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7917, 3.1427, 2.0345, 3.6191, 2.3219, 3.0984, 1.6914, 1.9248], + device='cuda:1'), covar=tensor([0.0236, 0.0497, 0.1834, 0.0356, 0.0330, 0.0414, 0.2636, 0.1434], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0224, 0.0293, 0.0233, 0.0242, 0.0242, 0.0265, 0.0282], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-20 23:14:29,416 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.498e+02 2.119e+02 2.439e+02 2.967e+02 4.703e+02, threshold=4.878e+02, percent-clipped=0.0 +2023-03-20 23:14:29,444 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 23:14:30,094 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40218.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:14:35,541 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8357, 3.3379, 3.8845, 3.7863, 3.7221, 3.9400, 3.6833, 3.5900], + device='cuda:1'), covar=tensor([0.0027, 0.0083, 0.0028, 0.0030, 0.0033, 0.0024, 0.0040, 0.0048], + device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0044, 0.0040, 0.0037, 0.0038, 0.0039, 0.0042, 0.0049], + device='cuda:1'), out_proj_covar=tensor([7.4791e-05, 1.1673e-04, 1.0313e-04, 8.6242e-05, 9.0159e-05, 9.3415e-05, + 1.0912e-04, 1.1756e-04], device='cuda:1') +2023-03-20 23:14:37,438 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 23:14:39,383 INFO [train.py:901] (1/2) Epoch 15, batch 700, loss[loss=0.1598, simple_loss=0.236, pruned_loss=0.04179, over 7239.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.239, pruned_loss=0.04529, over 1398070.12 frames. ], batch size: 89, lr: 1.10e-02, grad_scale: 8.0 +2023-03-20 23:14:49,025 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40256.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:14:57,148 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40272.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:15:00,615 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40279.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:15:00,978 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 23:15:01,954 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 23:15:04,419 INFO [train.py:901] (1/2) Epoch 15, batch 750, loss[loss=0.1751, simple_loss=0.2491, pruned_loss=0.05053, over 7304.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2381, pruned_loss=0.04506, over 1405309.61 frames. ], batch size: 80, lr: 1.10e-02, grad_scale: 8.0 +2023-03-20 23:15:09,084 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40295.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 23:15:17,612 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 23:15:20,573 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.345e+02 2.287e+02 2.600e+02 3.173e+02 8.989e+02, threshold=5.199e+02, percent-clipped=3.0 +2023-03-20 23:15:22,317 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 23:15:27,568 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40330.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:15:29,136 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40333.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:15:29,490 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 23:15:30,522 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 23:15:31,022 INFO [train.py:901] (1/2) Epoch 15, batch 800, loss[loss=0.1549, simple_loss=0.2211, pruned_loss=0.04432, over 7227.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2389, pruned_loss=0.04539, over 1414308.64 frames. ], batch size: 45, lr: 1.10e-02, grad_scale: 8.0 +2023-03-20 23:15:33,203 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8989, 3.1289, 1.9441, 3.4032, 2.2850, 2.8737, 1.6428, 1.9348], + device='cuda:1'), covar=tensor([0.0337, 0.0679, 0.2255, 0.0506, 0.0526, 0.0414, 0.3032, 0.1696], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0229, 0.0300, 0.0235, 0.0243, 0.0245, 0.0268, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-20 23:15:34,121 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=40343.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 23:15:38,640 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8139, 2.9770, 3.7250, 3.7928, 3.7188, 3.8338, 3.6802, 3.5442], + device='cuda:1'), covar=tensor([0.0024, 0.0096, 0.0030, 0.0027, 0.0032, 0.0025, 0.0039, 0.0051], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0046, 0.0042, 0.0038, 0.0040, 0.0041, 0.0044, 0.0050], + device='cuda:1'), out_proj_covar=tensor([7.8194e-05, 1.2175e-04, 1.0754e-04, 8.9499e-05, 9.3558e-05, 9.7574e-05, + 1.1350e-04, 1.2240e-04], device='cuda:1') +2023-03-20 23:15:41,020 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 23:15:56,611 INFO [train.py:901] (1/2) Epoch 15, batch 850, loss[loss=0.1767, simple_loss=0.2445, pruned_loss=0.05448, over 7264.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2397, pruned_loss=0.04616, over 1421358.99 frames. ], batch size: 47, lr: 1.10e-02, grad_scale: 16.0 +2023-03-20 23:15:59,326 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40391.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:16:00,198 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 23:16:00,670 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 23:16:06,044 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 23:16:09,130 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3282, 3.2017, 2.8907, 3.2085, 2.3711, 2.5811, 3.3346, 2.4558], + device='cuda:1'), covar=tensor([0.0231, 0.0274, 0.0270, 0.0332, 0.0428, 0.0537, 0.0406, 0.0959], + device='cuda:1'), in_proj_covar=tensor([0.0297, 0.0297, 0.0249, 0.0314, 0.0302, 0.0296, 0.0309, 0.0291], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 23:16:09,466 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 23:16:12,434 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.364e+02 2.208e+02 2.627e+02 3.087e+02 7.163e+02, threshold=5.253e+02, percent-clipped=1.0 +2023-03-20 23:16:15,591 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40423.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:16:17,399 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 +2023-03-20 23:16:22,542 INFO [train.py:901] (1/2) Epoch 15, batch 900, loss[loss=0.192, simple_loss=0.2685, pruned_loss=0.05779, over 7137.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2401, pruned_loss=0.04631, over 1426862.20 frames. ], batch size: 98, lr: 1.10e-02, grad_scale: 16.0 +2023-03-20 23:16:39,498 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:16:39,538 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:16:47,449 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 23:16:48,941 INFO [train.py:901] (1/2) Epoch 15, batch 950, loss[loss=0.1603, simple_loss=0.2372, pruned_loss=0.04164, over 7281.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2405, pruned_loss=0.04632, over 1431725.95 frames. ], batch size: 77, lr: 1.10e-02, grad_scale: 16.0 +2023-03-20 23:16:51,556 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40492.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:16:54,533 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9227, 2.3325, 1.9815, 2.5917, 2.6744, 2.4532, 2.4401, 2.3416], + device='cuda:1'), covar=tensor([0.2200, 0.0841, 0.3180, 0.0477, 0.0095, 0.0078, 0.0165, 0.0196], + device='cuda:1'), in_proj_covar=tensor([0.0289, 0.0234, 0.0282, 0.0260, 0.0136, 0.0132, 0.0163, 0.0185], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:17:03,787 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.511e+02 2.192e+02 2.721e+02 3.328e+02 8.028e+02, threshold=5.442e+02, percent-clipped=2.0 +2023-03-20 23:17:04,877 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=40519.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:17:10,344 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 23:17:13,798 INFO [train.py:901] (1/2) Epoch 15, batch 1000, loss[loss=0.1564, simple_loss=0.239, pruned_loss=0.03696, over 7267.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2402, pruned_loss=0.04608, over 1435380.14 frames. ], batch size: 89, lr: 1.10e-02, grad_scale: 16.0 +2023-03-20 23:17:20,859 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40551.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:17:21,971 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40553.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:17:31,581 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 23:17:33,660 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40574.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:17:37,217 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1482, 1.1701, 1.5697, 1.4437, 1.4583, 1.7135, 1.3190, 1.4635], + device='cuda:1'), covar=tensor([0.1952, 0.2742, 0.0688, 0.1263, 0.1234, 0.1499, 0.1301, 0.2447], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0051, 0.0036, 0.0034, 0.0040, 0.0038, 0.0053, 0.0042], + device='cuda:1'), out_proj_covar=tensor([1.1163e-04, 1.2738e-04, 9.8202e-05, 9.8725e-05, 1.0976e-04, 1.0847e-04, + 1.3159e-04, 1.1460e-04], device='cuda:1') +2023-03-20 23:17:40,095 INFO [train.py:901] (1/2) Epoch 15, batch 1050, loss[loss=0.1499, simple_loss=0.2328, pruned_loss=0.03354, over 7276.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2402, pruned_loss=0.04601, over 1437713.11 frames. ], batch size: 70, lr: 1.10e-02, grad_scale: 16.0 +2023-03-20 23:17:52,557 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 23:17:55,535 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.580e+02 2.140e+02 2.605e+02 3.056e+02 5.614e+02, threshold=5.210e+02, percent-clipped=1.0 +2023-03-20 23:17:56,538 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 23:18:01,101 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40628.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:18:05,673 INFO [train.py:901] (1/2) Epoch 15, batch 1100, loss[loss=0.1632, simple_loss=0.2453, pruned_loss=0.0406, over 7314.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2397, pruned_loss=0.0454, over 1440245.23 frames. ], batch size: 83, lr: 1.10e-02, grad_scale: 16.0 +2023-03-20 23:18:14,010 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40651.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:18:25,382 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 23:18:25,395 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 23:18:31,622 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40686.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:18:32,054 INFO [train.py:901] (1/2) Epoch 15, batch 1150, loss[loss=0.167, simple_loss=0.2513, pruned_loss=0.04133, over 7324.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.24, pruned_loss=0.04551, over 1441703.36 frames. ], batch size: 59, lr: 1.10e-02, grad_scale: 16.0 +2023-03-20 23:18:38,561 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 23:18:38,575 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 23:18:44,721 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40712.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:18:47,054 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 2.199e+02 2.637e+02 2.987e+02 6.275e+02, threshold=5.274e+02, percent-clipped=1.0 +2023-03-20 23:18:58,323 INFO [train.py:901] (1/2) Epoch 15, batch 1200, loss[loss=0.1678, simple_loss=0.2428, pruned_loss=0.04636, over 7297.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.24, pruned_loss=0.04549, over 1441509.08 frames. ], batch size: 57, lr: 1.10e-02, grad_scale: 16.0 +2023-03-20 23:19:10,027 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 23:19:23,439 INFO [train.py:901] (1/2) Epoch 15, batch 1250, loss[loss=0.145, simple_loss=0.2117, pruned_loss=0.03912, over 7193.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2393, pruned_loss=0.04522, over 1443268.10 frames. ], batch size: 39, lr: 1.09e-02, grad_scale: 16.0 +2023-03-20 23:19:28,121 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3226, 2.5363, 2.0130, 2.4215, 2.3082, 1.9423, 2.4315, 2.2478], + device='cuda:1'), covar=tensor([0.0608, 0.0615, 0.1047, 0.1082, 0.0811, 0.0761, 0.1284, 0.1666], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0042, 0.0048, 0.0043, 0.0042, 0.0044, 0.0044, 0.0041], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:19:33,778 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 23:19:36,200 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-20 23:19:37,035 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5767, 2.9093, 2.3593, 2.6163, 2.7735, 2.1641, 2.8273, 2.6264], + device='cuda:1'), covar=tensor([0.0791, 0.0916, 0.1387, 0.1346, 0.0805, 0.0936, 0.1870, 0.1179], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0042, 0.0048, 0.0043, 0.0042, 0.0044, 0.0044, 0.0040], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:19:38,411 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 23:19:39,342 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.392e+02 2.209e+02 2.625e+02 3.137e+02 7.215e+02, threshold=5.251e+02, percent-clipped=3.0 +2023-03-20 23:19:39,891 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 23:19:49,975 INFO [train.py:901] (1/2) Epoch 15, batch 1300, loss[loss=0.1922, simple_loss=0.2758, pruned_loss=0.05436, over 6790.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2399, pruned_loss=0.0458, over 1442718.46 frames. ], batch size: 106, lr: 1.09e-02, grad_scale: 16.0 +2023-03-20 23:19:55,484 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40848.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:19:56,971 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40851.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:20:03,481 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 23:20:03,656 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9563, 2.5713, 2.1999, 3.7601, 1.6064, 3.6096, 1.5198, 2.5673], + device='cuda:1'), covar=tensor([0.0067, 0.0806, 0.1635, 0.0082, 0.3754, 0.0100, 0.0971, 0.0170], + device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0256, 0.0302, 0.0158, 0.0283, 0.0164, 0.0265, 0.0211], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 23:20:04,110 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40865.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:20:05,506 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 23:20:08,552 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40874.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:20:08,981 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 23:20:14,950 INFO [train.py:901] (1/2) Epoch 15, batch 1350, loss[loss=0.1297, simple_loss=0.181, pruned_loss=0.03913, over 5993.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2394, pruned_loss=0.04568, over 1442784.54 frames. ], batch size: 25, lr: 1.09e-02, grad_scale: 16.0 +2023-03-20 23:20:15,713 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-20 23:20:20,023 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 23:20:21,738 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=40899.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:20:22,308 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40900.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:20:28,857 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5140, 3.6320, 2.4045, 4.1511, 3.0145, 3.3949, 2.0845, 2.2165], + device='cuda:1'), covar=tensor([0.0259, 0.0610, 0.1751, 0.0339, 0.0339, 0.0495, 0.2361, 0.1699], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0229, 0.0300, 0.0237, 0.0242, 0.0245, 0.0266, 0.0284], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-20 23:20:31,277 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.585e+02 2.254e+02 2.662e+02 3.076e+02 6.866e+02, threshold=5.325e+02, percent-clipped=2.0 +2023-03-20 23:20:33,824 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=40922.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:20:35,936 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40926.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:20:36,905 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40928.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:20:38,229 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-03-20 23:20:40,945 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.8451, 4.4601, 4.2911, 4.8815, 4.8112, 4.8456, 3.9307, 4.4144], + device='cuda:1'), covar=tensor([0.0825, 0.2419, 0.1803, 0.0857, 0.0657, 0.1300, 0.0924, 0.1035], + device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0309, 0.0247, 0.0235, 0.0182, 0.0301, 0.0169, 0.0214], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:20:41,367 INFO [train.py:901] (1/2) Epoch 15, batch 1400, loss[loss=0.1682, simple_loss=0.2466, pruned_loss=0.0449, over 7351.00 frames. ], tot_loss[loss=0.165, simple_loss=0.239, pruned_loss=0.04548, over 1442230.17 frames. ], batch size: 63, lr: 1.09e-02, grad_scale: 16.0 +2023-03-20 23:20:46,497 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2380, 3.3352, 2.0922, 3.5903, 2.6845, 3.1864, 1.6989, 1.8557], + device='cuda:1'), covar=tensor([0.0252, 0.0559, 0.2425, 0.0465, 0.0407, 0.0541, 0.2871, 0.1876], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0231, 0.0299, 0.0237, 0.0243, 0.0246, 0.0267, 0.0285], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-20 23:20:46,998 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0963, 3.3408, 2.2109, 3.8979, 2.7190, 3.0192, 1.7452, 1.9684], + device='cuda:1'), covar=tensor([0.0176, 0.0566, 0.1807, 0.0324, 0.0330, 0.0473, 0.2583, 0.1657], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0231, 0.0300, 0.0237, 0.0243, 0.0246, 0.0268, 0.0285], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-20 23:20:52,374 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 23:20:53,516 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40961.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 23:20:54,504 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6270, 2.8748, 3.5653, 3.7077, 3.6235, 3.7392, 3.5140, 3.2937], + device='cuda:1'), covar=tensor([0.0039, 0.0148, 0.0051, 0.0041, 0.0044, 0.0035, 0.0055, 0.0082], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0046, 0.0041, 0.0039, 0.0039, 0.0041, 0.0044, 0.0050], + device='cuda:1'), out_proj_covar=tensor([7.8813e-05, 1.2080e-04, 1.0601e-04, 9.0602e-05, 9.0639e-05, 9.4517e-05, + 1.1352e-04, 1.2118e-04], device='cuda:1') +2023-03-20 23:21:00,881 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=40976.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:21:06,541 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40986.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:21:06,949 INFO [train.py:901] (1/2) Epoch 15, batch 1450, loss[loss=0.1744, simple_loss=0.2413, pruned_loss=0.05376, over 7264.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2386, pruned_loss=0.04562, over 1441255.60 frames. ], batch size: 52, lr: 1.09e-02, grad_scale: 16.0 +2023-03-20 23:21:12,507 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40997.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:21:17,754 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 23:21:17,828 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41007.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:21:20,623 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 +2023-03-20 23:21:22,718 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.574e+02 2.243e+02 2.788e+02 3.650e+02 1.062e+03, threshold=5.577e+02, percent-clipped=7.0 +2023-03-20 23:21:31,131 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=41034.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:21:32,502 INFO [train.py:901] (1/2) Epoch 15, batch 1500, loss[loss=0.1543, simple_loss=0.2235, pruned_loss=0.0425, over 7149.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2386, pruned_loss=0.04553, over 1441309.66 frames. ], batch size: 41, lr: 1.09e-02, grad_scale: 16.0 +2023-03-20 23:21:33,521 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 23:21:36,712 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41045.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:21:41,778 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8252, 2.3460, 2.8110, 2.7814, 2.7132, 2.7241, 2.2255, 2.9775], + device='cuda:1'), covar=tensor([0.1329, 0.0480, 0.1150, 0.1796, 0.1036, 0.0820, 0.2921, 0.1109], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0041, 0.0036, 0.0037, 0.0035, 0.0032, 0.0050, 0.0037], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:21:43,254 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41058.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:21:57,278 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 23:21:58,767 INFO [train.py:901] (1/2) Epoch 15, batch 1550, loss[loss=0.1543, simple_loss=0.2337, pruned_loss=0.0375, over 7322.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.239, pruned_loss=0.04555, over 1443008.69 frames. ], batch size: 75, lr: 1.09e-02, grad_scale: 16.0 +2023-03-20 23:22:06,411 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41102.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:22:08,368 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41106.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:22:13,692 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.410e+02 2.850e+02 3.373e+02 6.504e+02, threshold=5.701e+02, percent-clipped=1.0 +2023-03-20 23:22:23,642 INFO [train.py:901] (1/2) Epoch 15, batch 1600, loss[loss=0.1589, simple_loss=0.2348, pruned_loss=0.04153, over 7322.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.239, pruned_loss=0.04563, over 1443961.44 frames. ], batch size: 59, lr: 1.09e-02, grad_scale: 16.0 +2023-03-20 23:22:24,814 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8894, 3.5764, 3.5286, 3.6560, 2.7889, 2.7443, 3.9132, 2.8951], + device='cuda:1'), covar=tensor([0.0197, 0.0285, 0.0240, 0.0262, 0.0375, 0.0552, 0.0329, 0.0888], + device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0304, 0.0247, 0.0315, 0.0302, 0.0295, 0.0305, 0.0289], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 23:22:26,655 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 23:22:27,634 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 23:22:29,177 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41148.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:22:30,073 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 23:22:35,474 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-03-20 23:22:37,237 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41163.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:22:40,742 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 23:22:45,186 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 23:22:45,501 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.90 vs. limit=2.0 +2023-03-20 23:22:49,584 INFO [train.py:901] (1/2) Epoch 15, batch 1650, loss[loss=0.1732, simple_loss=0.244, pruned_loss=0.05117, over 7331.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2399, pruned_loss=0.04595, over 1445109.26 frames. ], batch size: 61, lr: 1.09e-02, grad_scale: 16.0 +2023-03-20 23:22:52,672 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 23:22:54,266 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=41196.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:22:55,304 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4522, 1.9877, 2.2093, 1.5867, 1.8960, 1.3188, 1.5879, 1.3311], + device='cuda:1'), covar=tensor([0.0416, 0.0197, 0.0079, 0.0153, 0.0443, 0.0355, 0.0365, 0.0315], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0023, 0.0022, 0.0020, 0.0022, 0.0021, 0.0024, 0.0026], + device='cuda:1'), out_proj_covar=tensor([5.9744e-05, 5.7589e-05, 5.3297e-05, 4.8982e-05, 5.6967e-05, 5.4549e-05, + 5.8024e-05, 6.4370e-05], device='cuda:1') +2023-03-20 23:23:04,927 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 2.271e+02 2.648e+02 3.290e+02 5.745e+02, threshold=5.296e+02, percent-clipped=1.0 +2023-03-20 23:23:07,087 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41221.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:23:10,563 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 23:23:15,224 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 23:23:15,709 INFO [train.py:901] (1/2) Epoch 15, batch 1700, loss[loss=0.1521, simple_loss=0.2277, pruned_loss=0.03832, over 7267.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2393, pruned_loss=0.04579, over 1442620.13 frames. ], batch size: 52, lr: 1.09e-02, grad_scale: 16.0 +2023-03-20 23:23:25,903 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41256.0, num_to_drop=1, layers_to_drop={3} +2023-03-20 23:23:26,293 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 23:23:39,685 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 +2023-03-20 23:23:41,293 INFO [train.py:901] (1/2) Epoch 15, batch 1750, loss[loss=0.2016, simple_loss=0.2692, pruned_loss=0.06702, over 6668.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2395, pruned_loss=0.04576, over 1441780.11 frames. ], batch size: 106, lr: 1.09e-02, grad_scale: 16.0 +2023-03-20 23:23:43,440 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2407, 3.2471, 2.1203, 3.7603, 2.4226, 3.1568, 1.7336, 1.9509], + device='cuda:1'), covar=tensor([0.0274, 0.0570, 0.2037, 0.0405, 0.0277, 0.0631, 0.2689, 0.1735], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0232, 0.0305, 0.0240, 0.0241, 0.0250, 0.0270, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-20 23:23:46,439 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5233, 2.0189, 2.2127, 1.5398, 1.9845, 1.3090, 1.6310, 1.2919], + device='cuda:1'), covar=tensor([0.0675, 0.0268, 0.0107, 0.0219, 0.0442, 0.0511, 0.0420, 0.0554], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0023, 0.0022, 0.0020, 0.0022, 0.0021, 0.0024, 0.0025], + device='cuda:1'), out_proj_covar=tensor([5.9189e-05, 5.6869e-05, 5.2296e-05, 4.8410e-05, 5.5721e-05, 5.3595e-05, + 5.6773e-05, 6.2403e-05], device='cuda:1') +2023-03-20 23:23:50,773 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 23:23:51,389 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41307.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:23:51,775 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 23:23:56,292 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.319e+02 2.224e+02 2.586e+02 3.166e+02 5.142e+02, threshold=5.173e+02, percent-clipped=0.0 +2023-03-20 23:24:05,669 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7519, 3.6337, 2.4456, 4.1810, 3.1993, 3.4423, 2.0677, 2.2484], + device='cuda:1'), covar=tensor([0.0227, 0.0560, 0.1980, 0.0444, 0.0362, 0.0486, 0.2466, 0.2139], + device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0233, 0.0308, 0.0243, 0.0246, 0.0251, 0.0273, 0.0296], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-20 23:24:06,646 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1527, 4.1915, 3.6724, 3.5726, 3.5950, 2.4252, 1.8312, 4.0373], + device='cuda:1'), covar=tensor([0.0024, 0.0027, 0.0063, 0.0048, 0.0064, 0.0358, 0.0503, 0.0040], + device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0067, 0.0087, 0.0072, 0.0094, 0.0111, 0.0112, 0.0078], + device='cuda:1'), out_proj_covar=tensor([9.6474e-05, 9.6712e-05, 1.1486e-04, 1.0038e-04, 1.2282e-04, 1.4660e-04, + 1.4894e-04, 1.0190e-04], device='cuda:1') +2023-03-20 23:24:07,016 INFO [train.py:901] (1/2) Epoch 15, batch 1800, loss[loss=0.1827, simple_loss=0.2553, pruned_loss=0.05507, over 7248.00 frames. ], tot_loss[loss=0.165, simple_loss=0.239, pruned_loss=0.04552, over 1441110.99 frames. ], batch size: 64, lr: 1.09e-02, grad_scale: 16.0 +2023-03-20 23:24:13,997 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 23:24:15,593 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41353.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:24:16,569 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=41355.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:24:26,489 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 23:24:27,132 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0903, 1.3278, 1.2720, 1.1710, 1.2880, 1.1127, 1.0715, 0.8275], + device='cuda:1'), covar=tensor([0.0120, 0.0072, 0.0136, 0.0094, 0.0134, 0.0106, 0.0096, 0.0131], + device='cuda:1'), in_proj_covar=tensor([0.0023, 0.0021, 0.0022, 0.0021, 0.0023, 0.0021, 0.0024, 0.0028], + device='cuda:1'), out_proj_covar=tensor([2.7369e-05, 2.4527e-05, 2.5842e-05, 2.3972e-05, 2.7745e-05, 2.3575e-05, + 2.7266e-05, 3.3595e-05], device='cuda:1') +2023-03-20 23:24:32,444 INFO [train.py:901] (1/2) Epoch 15, batch 1850, loss[loss=0.1422, simple_loss=0.2269, pruned_loss=0.02877, over 7301.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2396, pruned_loss=0.04569, over 1441788.14 frames. ], batch size: 68, lr: 1.09e-02, grad_scale: 16.0 +2023-03-20 23:24:36,490 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 23:24:39,892 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41401.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:24:48,353 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.233e+02 2.216e+02 2.442e+02 3.300e+02 6.714e+02, threshold=4.883e+02, percent-clipped=2.0 +2023-03-20 23:24:53,021 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 23:24:58,914 INFO [train.py:901] (1/2) Epoch 15, batch 1900, loss[loss=0.1812, simple_loss=0.2494, pruned_loss=0.05651, over 7274.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2394, pruned_loss=0.04594, over 1439866.17 frames. ], batch size: 70, lr: 1.09e-02, grad_scale: 16.0 +2023-03-20 23:25:09,607 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41458.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:25:18,216 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 23:25:24,041 INFO [train.py:901] (1/2) Epoch 15, batch 1950, loss[loss=0.1658, simple_loss=0.2379, pruned_loss=0.04687, over 7345.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2386, pruned_loss=0.04561, over 1439706.89 frames. ], batch size: 61, lr: 1.09e-02, grad_scale: 16.0 +2023-03-20 23:25:29,033 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 23:25:33,653 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 23:25:34,626 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 23:25:40,161 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.628e+02 2.202e+02 2.600e+02 3.272e+02 6.051e+02, threshold=5.201e+02, percent-clipped=3.0 +2023-03-20 23:25:42,283 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41521.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:25:50,211 INFO [train.py:901] (1/2) Epoch 15, batch 2000, loss[loss=0.1743, simple_loss=0.243, pruned_loss=0.0528, over 7266.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2388, pruned_loss=0.04567, over 1442121.41 frames. ], batch size: 70, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:25:51,717 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 23:25:59,863 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41556.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 23:26:02,217 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 23:26:05,293 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5896, 3.2560, 3.3290, 3.6026, 3.5466, 3.5810, 3.2846, 3.4221], + device='cuda:1'), covar=tensor([0.0028, 0.0073, 0.0046, 0.0031, 0.0033, 0.0035, 0.0058, 0.0055], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0047, 0.0042, 0.0040, 0.0041, 0.0042, 0.0045, 0.0052], + device='cuda:1'), out_proj_covar=tensor([7.8663e-05, 1.2375e-04, 1.0573e-04, 9.4461e-05, 9.3434e-05, 9.7505e-05, + 1.1511e-04, 1.2472e-04], device='cuda:1') +2023-03-20 23:26:06,250 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=41569.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:26:09,652 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 23:26:15,762 INFO [train.py:901] (1/2) Epoch 15, batch 2050, loss[loss=0.1576, simple_loss=0.2364, pruned_loss=0.03941, over 7309.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2395, pruned_loss=0.04558, over 1445977.00 frames. ], batch size: 83, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:26:25,274 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=41604.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:26:31,477 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2641, 1.9346, 2.3052, 1.5192, 2.0031, 1.3836, 1.5965, 1.3715], + device='cuda:1'), covar=tensor([0.0438, 0.0268, 0.0086, 0.0139, 0.0203, 0.0420, 0.0290, 0.0307], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0023, 0.0022, 0.0020, 0.0021, 0.0021, 0.0023, 0.0025], + device='cuda:1'), out_proj_covar=tensor([5.9952e-05, 5.8475e-05, 5.2974e-05, 4.8781e-05, 5.4850e-05, 5.4315e-05, + 5.6512e-05, 6.2366e-05], device='cuda:1') +2023-03-20 23:26:31,810 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.491e+02 2.163e+02 2.642e+02 3.130e+02 5.722e+02, threshold=5.285e+02, percent-clipped=2.0 +2023-03-20 23:26:41,953 INFO [train.py:901] (1/2) Epoch 15, batch 2100, loss[loss=0.1755, simple_loss=0.254, pruned_loss=0.04844, over 7346.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2399, pruned_loss=0.0456, over 1445418.61 frames. ], batch size: 63, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:26:42,092 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1672, 1.1570, 1.6247, 1.6620, 1.4067, 1.5874, 1.3875, 1.5181], + device='cuda:1'), covar=tensor([0.1663, 0.3539, 0.0611, 0.1121, 0.1907, 0.2847, 0.1310, 0.2030], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0052, 0.0037, 0.0035, 0.0040, 0.0038, 0.0055, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 23:26:44,497 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 23:26:47,418 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 23:26:49,972 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41653.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:26:50,510 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41654.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:27:00,579 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-20 23:27:07,956 INFO [train.py:901] (1/2) Epoch 15, batch 2150, loss[loss=0.1888, simple_loss=0.257, pruned_loss=0.06031, over 7250.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2404, pruned_loss=0.04568, over 1445934.20 frames. ], batch size: 89, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:27:15,088 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:27:15,150 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:27:22,203 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41715.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 23:27:23,071 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.084e+02 2.496e+02 3.208e+02 6.072e+02, threshold=4.993e+02, percent-clipped=2.0 +2023-03-20 23:27:33,050 INFO [train.py:901] (1/2) Epoch 15, batch 2200, loss[loss=0.164, simple_loss=0.236, pruned_loss=0.04595, over 7331.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2393, pruned_loss=0.04547, over 1445856.21 frames. ], batch size: 54, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:27:35,150 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 23:27:39,265 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=41749.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:27:43,769 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41758.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:27:47,574 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-03-20 23:27:54,039 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7716, 3.0714, 2.4200, 2.8922, 2.8825, 2.4030, 2.9524, 2.8350], + device='cuda:1'), covar=tensor([0.1362, 0.0666, 0.1198, 0.1418, 0.0946, 0.1053, 0.1198, 0.1487], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0041, 0.0047, 0.0041, 0.0041, 0.0043, 0.0044, 0.0040], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:27:59,461 INFO [train.py:901] (1/2) Epoch 15, batch 2250, loss[loss=0.1575, simple_loss=0.2348, pruned_loss=0.04004, over 7322.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2388, pruned_loss=0.04528, over 1444535.40 frames. ], batch size: 83, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:28:00,101 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0975, 3.7633, 4.0441, 4.0908, 4.1765, 4.1653, 4.2161, 4.0271], + device='cuda:1'), covar=tensor([0.0025, 0.0070, 0.0031, 0.0029, 0.0025, 0.0025, 0.0024, 0.0038], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0047, 0.0042, 0.0040, 0.0040, 0.0042, 0.0045, 0.0052], + device='cuda:1'), out_proj_covar=tensor([7.7596e-05, 1.2367e-04, 1.0576e-04, 9.1870e-05, 9.2359e-05, 9.8172e-05, + 1.1489e-04, 1.2385e-04], device='cuda:1') +2023-03-20 23:28:09,410 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=41806.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:28:10,391 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 23:28:10,877 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 23:28:14,906 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.381e+02 2.118e+02 2.612e+02 3.099e+02 6.115e+02, threshold=5.224e+02, percent-clipped=1.0 +2023-03-20 23:28:23,431 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 23:28:24,927 INFO [train.py:901] (1/2) Epoch 15, batch 2300, loss[loss=0.1834, simple_loss=0.2583, pruned_loss=0.05421, over 7357.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2394, pruned_loss=0.04556, over 1444406.66 frames. ], batch size: 73, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:28:37,671 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6009, 2.4488, 2.2365, 3.8494, 1.7608, 3.4070, 1.3459, 2.9261], + device='cuda:1'), covar=tensor([0.0064, 0.0748, 0.1426, 0.0058, 0.3062, 0.0089, 0.0986, 0.0206], + device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0253, 0.0294, 0.0152, 0.0279, 0.0165, 0.0263, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 23:28:51,141 INFO [train.py:901] (1/2) Epoch 15, batch 2350, loss[loss=0.1602, simple_loss=0.2345, pruned_loss=0.04294, over 7282.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2387, pruned_loss=0.04506, over 1443661.58 frames. ], batch size: 66, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:28:51,790 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4451, 1.7558, 2.0709, 1.6127, 1.7987, 1.4973, 1.6363, 1.2744], + device='cuda:1'), covar=tensor([0.0259, 0.0187, 0.0060, 0.0129, 0.0205, 0.0242, 0.0222, 0.0194], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0024, 0.0022, 0.0021, 0.0022, 0.0022, 0.0024, 0.0025], + device='cuda:1'), out_proj_covar=tensor([6.1342e-05, 5.9631e-05, 5.3161e-05, 5.0393e-05, 5.7069e-05, 5.5146e-05, + 5.8767e-05, 6.3465e-05], device='cuda:1') +2023-03-20 23:29:01,010 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41906.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:29:06,454 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.373e+02 2.041e+02 2.774e+02 3.527e+02 1.027e+03, threshold=5.547e+02, percent-clipped=7.0 +2023-03-20 23:29:08,992 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 23:29:10,686 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6140, 3.1640, 2.5675, 4.1890, 1.8352, 3.6244, 1.4998, 3.1576], + device='cuda:1'), covar=tensor([0.0082, 0.0580, 0.1488, 0.0051, 0.3689, 0.0114, 0.1061, 0.0253], + device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0255, 0.0293, 0.0151, 0.0280, 0.0165, 0.0261, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 23:29:16,032 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 23:29:16,497 INFO [train.py:901] (1/2) Epoch 15, batch 2400, loss[loss=0.1533, simple_loss=0.2278, pruned_loss=0.03937, over 7287.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2382, pruned_loss=0.04488, over 1443483.80 frames. ], batch size: 86, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:29:22,856 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41948.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 23:29:26,412 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8007, 3.4362, 3.6660, 3.5054, 3.4495, 3.4387, 3.7338, 3.2499], + device='cuda:1'), covar=tensor([0.0139, 0.0195, 0.0107, 0.0162, 0.0365, 0.0125, 0.0148, 0.0195], + device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0073, 0.0073, 0.0063, 0.0126, 0.0082, 0.0078, 0.0077], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:29:27,850 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 23:29:30,884 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 23:29:33,002 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41967.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:29:42,899 INFO [train.py:901] (1/2) Epoch 15, batch 2450, loss[loss=0.1791, simple_loss=0.2571, pruned_loss=0.05061, over 7265.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2387, pruned_loss=0.04495, over 1445106.92 frames. ], batch size: 89, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:29:54,513 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42009.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 23:29:54,933 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42010.0, num_to_drop=1, layers_to_drop={0} +2023-03-20 23:29:58,412 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.325e+02 2.135e+02 2.470e+02 3.053e+02 4.844e+02, threshold=4.939e+02, percent-clipped=0.0 +2023-03-20 23:29:58,432 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 23:30:09,675 INFO [train.py:901] (1/2) Epoch 15, batch 2500, loss[loss=0.1869, simple_loss=0.2623, pruned_loss=0.05577, over 6759.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2389, pruned_loss=0.04515, over 1445088.99 frames. ], batch size: 106, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:30:23,671 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 23:30:34,628 INFO [train.py:901] (1/2) Epoch 15, batch 2550, loss[loss=0.1624, simple_loss=0.2412, pruned_loss=0.04181, over 7319.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2386, pruned_loss=0.04507, over 1443916.93 frames. ], batch size: 83, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:30:35,319 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2298, 2.3638, 2.1409, 3.7332, 1.5278, 3.1766, 1.2375, 2.7423], + device='cuda:1'), covar=tensor([0.0066, 0.0860, 0.1595, 0.0065, 0.3919, 0.0101, 0.1118, 0.0157], + device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0254, 0.0296, 0.0152, 0.0283, 0.0165, 0.0265, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 23:30:49,656 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 +2023-03-20 23:30:49,758 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.459e+02 2.206e+02 2.616e+02 3.137e+02 6.441e+02, threshold=5.231e+02, percent-clipped=2.0 +2023-03-20 23:30:56,801 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-20 23:30:58,159 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1449, 3.2910, 1.9744, 3.4585, 2.3991, 3.0407, 1.6938, 1.8079], + device='cuda:1'), covar=tensor([0.0245, 0.0572, 0.1885, 0.0469, 0.0329, 0.0337, 0.2651, 0.1645], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0230, 0.0299, 0.0239, 0.0240, 0.0239, 0.0262, 0.0283], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-20 23:31:00,587 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8380, 2.2813, 2.7865, 2.6537, 2.7430, 2.5260, 2.0965, 2.7382], + device='cuda:1'), covar=tensor([0.1476, 0.0749, 0.1436, 0.1747, 0.1537, 0.1283, 0.3890, 0.1484], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0044, 0.0037, 0.0039, 0.0036, 0.0034, 0.0052, 0.0039], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:31:00,958 INFO [train.py:901] (1/2) Epoch 15, batch 2600, loss[loss=0.157, simple_loss=0.2391, pruned_loss=0.03746, over 7295.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2388, pruned_loss=0.04482, over 1446928.34 frames. ], batch size: 86, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:31:18,556 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0686, 1.3903, 1.2646, 1.4118, 1.2611, 1.1907, 1.0525, 0.9185], + device='cuda:1'), covar=tensor([0.0152, 0.0084, 0.0181, 0.0135, 0.0107, 0.0091, 0.0238, 0.0153], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0021, 0.0022, 0.0022, 0.0022, 0.0021, 0.0024, 0.0028], + device='cuda:1'), out_proj_covar=tensor([2.7988e-05, 2.4297e-05, 2.5747e-05, 2.4376e-05, 2.7121e-05, 2.3688e-05, + 2.7002e-05, 3.3883e-05], device='cuda:1') +2023-03-20 23:31:25,860 INFO [train.py:901] (1/2) Epoch 15, batch 2650, loss[loss=0.1279, simple_loss=0.2063, pruned_loss=0.02473, over 7314.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2386, pruned_loss=0.0447, over 1446540.50 frames. ], batch size: 44, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:31:41,014 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.117e+02 2.438e+02 2.904e+02 5.048e+02, threshold=4.876e+02, percent-clipped=0.0 +2023-03-20 23:31:50,884 INFO [train.py:901] (1/2) Epoch 15, batch 2700, loss[loss=0.1765, simple_loss=0.2524, pruned_loss=0.0503, over 7348.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.238, pruned_loss=0.04448, over 1445038.53 frames. ], batch size: 63, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:32:03,249 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42262.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:32:14,827 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5584, 4.4187, 4.0622, 3.8904, 4.0182, 2.7020, 2.3302, 4.5102], + device='cuda:1'), covar=tensor([0.0027, 0.0077, 0.0072, 0.0054, 0.0056, 0.0358, 0.0458, 0.0033], + device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0067, 0.0089, 0.0074, 0.0096, 0.0112, 0.0114, 0.0079], + device='cuda:1'), out_proj_covar=tensor([9.8565e-05, 9.6724e-05, 1.1831e-04, 1.0200e-04, 1.2497e-04, 1.4709e-04, + 1.5008e-04, 1.0283e-04], device='cuda:1') +2023-03-20 23:32:15,715 INFO [train.py:901] (1/2) Epoch 15, batch 2750, loss[loss=0.151, simple_loss=0.2197, pruned_loss=0.04117, over 7262.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2371, pruned_loss=0.04428, over 1441936.61 frames. ], batch size: 47, lr: 1.08e-02, grad_scale: 16.0 +2023-03-20 23:32:24,304 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42304.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 23:32:27,294 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42310.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 23:32:30,644 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.441e+02 2.237e+02 2.733e+02 3.339e+02 6.919e+02, threshold=5.467e+02, percent-clipped=1.0 +2023-03-20 23:32:40,400 INFO [train.py:901] (1/2) Epoch 15, batch 2800, loss[loss=0.2092, simple_loss=0.277, pruned_loss=0.07064, over 6775.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2385, pruned_loss=0.04473, over 1442705.09 frames. ], batch size: 107, lr: 1.07e-02, grad_scale: 16.0 +2023-03-20 23:32:46,974 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 23:32:50,601 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=42358.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:33:03,623 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 23:33:03,965 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 23:33:03,987 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 23:33:04,005 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 23:33:04,177 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 23:33:04,192 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 23:33:04,203 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 23:33:04,304 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 23:33:04,379 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 23:33:04,393 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 23:33:04,411 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 23:33:04,575 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 23:33:04,581 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 23:33:04,589 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 23:33:05,060 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 23:33:05,063 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 23:33:05,119 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 23:33:13,489 INFO [train.py:901] (1/2) Epoch 16, batch 0, loss[loss=0.1752, simple_loss=0.2494, pruned_loss=0.05049, over 6871.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2494, pruned_loss=0.05049, over 6871.00 frames. ], batch size: 107, lr: 1.04e-02, grad_scale: 32.0 +2023-03-20 23:33:13,489 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 23:33:25,452 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8163, 3.4062, 3.5104, 3.5536, 3.5509, 3.4818, 3.6456, 3.3444], + device='cuda:1'), covar=tensor([0.0082, 0.0148, 0.0115, 0.0110, 0.0270, 0.0095, 0.0126, 0.0147], + device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0073, 0.0073, 0.0063, 0.0127, 0.0082, 0.0079, 0.0078], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:33:37,711 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3311, 1.8502, 2.2302, 1.7730, 2.0022, 1.5488, 1.8953, 1.5781], + device='cuda:1'), covar=tensor([0.0569, 0.0495, 0.0148, 0.0167, 0.0465, 0.0443, 0.0217, 0.0185], + device='cuda:1'), in_proj_covar=tensor([0.0023, 0.0023, 0.0022, 0.0020, 0.0020, 0.0020, 0.0023, 0.0023], + device='cuda:1'), out_proj_covar=tensor([5.8221e-05, 5.7401e-05, 5.1984e-05, 4.8366e-05, 5.3618e-05, 5.2072e-05, + 5.5183e-05, 5.9053e-05], device='cuda:1') +2023-03-20 23:33:39,696 INFO [train.py:935] (1/2) Epoch 16, validation: loss=0.1677, simple_loss=0.2552, pruned_loss=0.04012, over 1622729.00 frames. +2023-03-20 23:33:39,697 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-20 23:33:46,809 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 23:33:49,475 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42379.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:33:55,683 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-20 23:33:57,847 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 23:34:05,302 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 23:34:05,821 INFO [train.py:901] (1/2) Epoch 16, batch 50, loss[loss=0.1631, simple_loss=0.2364, pruned_loss=0.04492, over 7311.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.238, pruned_loss=0.0435, over 326299.12 frames. ], batch size: 59, lr: 1.04e-02, grad_scale: 16.0 +2023-03-20 23:34:07,364 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 23:34:09,365 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.463e+02 2.229e+02 2.517e+02 3.017e+02 7.171e+02, threshold=5.034e+02, percent-clipped=2.0 +2023-03-20 23:34:10,396 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 23:34:20,554 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42440.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:34:28,166 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 23:34:28,628 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 23:34:31,681 INFO [train.py:901] (1/2) Epoch 16, batch 100, loss[loss=0.1741, simple_loss=0.2563, pruned_loss=0.04594, over 6594.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2363, pruned_loss=0.04278, over 572176.87 frames. ], batch size: 106, lr: 1.04e-02, grad_scale: 16.0 +2023-03-20 23:34:55,146 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42506.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:34:57,508 INFO [train.py:901] (1/2) Epoch 16, batch 150, loss[loss=0.1454, simple_loss=0.2276, pruned_loss=0.03159, over 7287.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2376, pruned_loss=0.04434, over 764259.36 frames. ], batch size: 68, lr: 1.04e-02, grad_scale: 16.0 +2023-03-20 23:35:01,030 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.394e+02 2.163e+02 2.502e+02 3.239e+02 5.751e+02, threshold=5.003e+02, percent-clipped=1.0 +2023-03-20 23:35:23,797 INFO [train.py:901] (1/2) Epoch 16, batch 200, loss[loss=0.1582, simple_loss=0.233, pruned_loss=0.04172, over 7266.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2374, pruned_loss=0.04461, over 915530.25 frames. ], batch size: 77, lr: 1.04e-02, grad_scale: 16.0 +2023-03-20 23:35:24,401 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42562.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:35:26,931 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42567.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:35:29,851 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-20 23:35:34,905 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-20 23:35:40,880 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-20 23:35:44,958 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 +2023-03-20 23:35:45,755 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42604.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 23:35:46,210 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6539, 4.2567, 4.2659, 4.7602, 4.7380, 4.6759, 3.9652, 4.1780], + device='cuda:1'), covar=tensor([0.0740, 0.2596, 0.1809, 0.0937, 0.0736, 0.1352, 0.0841, 0.1192], + device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0312, 0.0248, 0.0242, 0.0182, 0.0298, 0.0170, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:35:48,711 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=42610.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:35:48,795 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3732, 1.6642, 2.2458, 1.6900, 1.7114, 1.6510, 1.7584, 1.4909], + device='cuda:1'), covar=tensor([0.0374, 0.0369, 0.0088, 0.0142, 0.0533, 0.0523, 0.0273, 0.0191], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0023, 0.0021, 0.0021, 0.0021, 0.0021, 0.0024, 0.0024], + device='cuda:1'), out_proj_covar=tensor([5.9894e-05, 5.8307e-05, 5.2394e-05, 4.9528e-05, 5.4546e-05, 5.2843e-05, + 5.7141e-05, 6.0501e-05], device='cuda:1') +2023-03-20 23:35:49,156 INFO [train.py:901] (1/2) Epoch 16, batch 250, loss[loss=0.1494, simple_loss=0.2236, pruned_loss=0.03758, over 7275.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2369, pruned_loss=0.04376, over 1033952.09 frames. ], batch size: 47, lr: 1.04e-02, grad_scale: 16.0 +2023-03-20 23:35:52,732 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.263e+02 2.075e+02 2.482e+02 2.956e+02 5.105e+02, threshold=4.965e+02, percent-clipped=1.0 +2023-03-20 23:35:54,749 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-20 23:36:09,596 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42649.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:36:11,029 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=42652.0, num_to_drop=1, layers_to_drop={1} +2023-03-20 23:36:15,467 INFO [train.py:901] (1/2) Epoch 16, batch 300, loss[loss=0.1723, simple_loss=0.2466, pruned_loss=0.04905, over 7340.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2377, pruned_loss=0.04433, over 1125490.69 frames. ], batch size: 61, lr: 1.04e-02, grad_scale: 16.0 +2023-03-20 23:36:15,996 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-20 23:36:24,932 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-20 23:36:27,971 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0945, 1.0642, 1.5456, 1.4729, 1.4001, 1.5504, 1.2255, 1.4090], + device='cuda:1'), covar=tensor([0.1714, 0.2515, 0.0704, 0.1117, 0.1053, 0.1571, 0.0718, 0.2002], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0051, 0.0036, 0.0035, 0.0040, 0.0038, 0.0056, 0.0040], + device='cuda:1'), out_proj_covar=tensor([1.1272e-04, 1.2822e-04, 9.9985e-05, 1.0304e-04, 1.1074e-04, 1.1022e-04, + 1.3877e-04, 1.1248e-04], device='cuda:1') +2023-03-20 23:36:40,084 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42710.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:36:40,455 INFO [train.py:901] (1/2) Epoch 16, batch 350, loss[loss=0.144, simple_loss=0.2195, pruned_loss=0.03419, over 7323.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2374, pruned_loss=0.044, over 1196157.38 frames. ], batch size: 75, lr: 1.04e-02, grad_scale: 16.0 +2023-03-20 23:36:44,530 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.340e+02 2.627e+02 3.104e+02 4.812e+02, threshold=5.255e+02, percent-clipped=0.0 +2023-03-20 23:36:46,375 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 +2023-03-20 23:36:53,832 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42735.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:36:58,869 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-20 23:37:06,935 INFO [train.py:901] (1/2) Epoch 16, batch 400, loss[loss=0.167, simple_loss=0.241, pruned_loss=0.0465, over 7360.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2378, pruned_loss=0.04399, over 1253061.48 frames. ], batch size: 73, lr: 1.04e-02, grad_scale: 16.0 +2023-03-20 23:37:31,849 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4276, 2.3796, 2.1076, 3.5546, 1.5207, 3.2127, 1.1905, 2.8454], + device='cuda:1'), covar=tensor([0.0088, 0.0882, 0.1599, 0.0059, 0.3803, 0.0089, 0.1094, 0.0210], + device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0259, 0.0296, 0.0154, 0.0284, 0.0166, 0.0267, 0.0215], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 23:37:33,186 INFO [train.py:901] (1/2) Epoch 16, batch 450, loss[loss=0.1508, simple_loss=0.2315, pruned_loss=0.03507, over 7252.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.237, pruned_loss=0.04379, over 1294228.65 frames. ], batch size: 55, lr: 1.04e-02, grad_scale: 16.0 +2023-03-20 23:37:37,217 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.066e+02 2.412e+02 3.102e+02 6.597e+02, threshold=4.823e+02, percent-clipped=2.0 +2023-03-20 23:37:40,771 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-20 23:37:40,784 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-20 23:37:57,568 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42859.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:37:58,448 INFO [train.py:901] (1/2) Epoch 16, batch 500, loss[loss=0.1799, simple_loss=0.2482, pruned_loss=0.0558, over 7230.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2374, pruned_loss=0.04384, over 1328590.54 frames. ], batch size: 50, lr: 1.04e-02, grad_scale: 8.0 +2023-03-20 23:37:59,071 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42862.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:38:00,141 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2086, 0.9580, 1.5841, 1.4082, 1.4269, 1.4417, 1.1823, 1.3974], + device='cuda:1'), covar=tensor([0.1028, 0.2916, 0.0943, 0.1074, 0.0892, 0.1322, 0.0688, 0.1139], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0053, 0.0038, 0.0037, 0.0040, 0.0039, 0.0056, 0.0041], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 23:38:07,829 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42879.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:38:08,045 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.03 vs. limit=2.0 +2023-03-20 23:38:13,810 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-20 23:38:15,303 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-20 23:38:16,263 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-20 23:38:17,529 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.75 vs. limit=5.0 +2023-03-20 23:38:18,263 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-20 23:38:22,995 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-20 23:38:24,947 INFO [train.py:901] (1/2) Epoch 16, batch 550, loss[loss=0.1646, simple_loss=0.2362, pruned_loss=0.04653, over 7245.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2376, pruned_loss=0.04418, over 1352136.45 frames. ], batch size: 45, lr: 1.04e-02, grad_scale: 8.0 +2023-03-20 23:38:28,881 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.372e+02 2.137e+02 2.590e+02 3.192e+02 4.525e+02, threshold=5.179e+02, percent-clipped=0.0 +2023-03-20 23:38:29,587 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42920.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:38:33,988 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-20 23:38:39,709 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42940.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:38:41,201 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4751, 1.8498, 2.0852, 1.8795, 2.0559, 1.5642, 1.7793, 1.4261], + device='cuda:1'), covar=tensor([0.0336, 0.0229, 0.0109, 0.0081, 0.0286, 0.0395, 0.0181, 0.0259], + device='cuda:1'), in_proj_covar=tensor([0.0023, 0.0022, 0.0021, 0.0020, 0.0020, 0.0019, 0.0023, 0.0023], + device='cuda:1'), out_proj_covar=tensor([5.8559e-05, 5.6021e-05, 5.0810e-05, 4.7718e-05, 5.2224e-05, 5.0387e-05, + 5.5087e-05, 5.9091e-05], device='cuda:1') +2023-03-20 23:38:42,589 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-20 23:38:45,594 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-20 23:38:50,135 INFO [train.py:901] (1/2) Epoch 16, batch 600, loss[loss=0.2241, simple_loss=0.2872, pruned_loss=0.08053, over 6733.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2367, pruned_loss=0.04404, over 1370512.68 frames. ], batch size: 107, lr: 1.04e-02, grad_scale: 8.0 +2023-03-20 23:38:53,029 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-20 23:39:09,159 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-20 23:39:14,035 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43005.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:39:17,008 INFO [train.py:901] (1/2) Epoch 16, batch 650, loss[loss=0.1509, simple_loss=0.2266, pruned_loss=0.03763, over 7233.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2362, pruned_loss=0.04365, over 1387117.21 frames. ], batch size: 45, lr: 1.04e-02, grad_scale: 8.0 +2023-03-20 23:39:18,030 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-20 23:39:21,011 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.149e+02 2.474e+02 3.215e+02 8.384e+02, threshold=4.948e+02, percent-clipped=2.0 +2023-03-20 23:39:21,153 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3055, 3.8668, 4.1757, 4.1403, 4.2603, 4.1663, 4.1794, 4.1338], + device='cuda:1'), covar=tensor([0.0025, 0.0061, 0.0027, 0.0024, 0.0024, 0.0025, 0.0025, 0.0035], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0048, 0.0044, 0.0040, 0.0042, 0.0044, 0.0045, 0.0052], + device='cuda:1'), out_proj_covar=tensor([8.1517e-05, 1.2432e-04, 1.1021e-04, 9.1375e-05, 9.5056e-05, 1.0154e-04, + 1.1420e-04, 1.2184e-04], device='cuda:1') +2023-03-20 23:39:21,897 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 +2023-03-20 23:39:29,249 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43035.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:39:35,706 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-20 23:39:42,908 INFO [train.py:901] (1/2) Epoch 16, batch 700, loss[loss=0.16, simple_loss=0.2315, pruned_loss=0.04422, over 7236.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2366, pruned_loss=0.04387, over 1400382.60 frames. ], batch size: 45, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:39:44,385 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-20 23:39:46,045 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9210, 2.8460, 2.7449, 2.6524, 2.1788, 2.4126, 2.8204, 2.2700], + device='cuda:1'), covar=tensor([0.0283, 0.0296, 0.0320, 0.0345, 0.0442, 0.0560, 0.0413, 0.1038], + device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0307, 0.0253, 0.0325, 0.0305, 0.0301, 0.0316, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 23:39:47,472 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1998, 4.7362, 4.7819, 4.7725, 4.6966, 4.2755, 4.8538, 4.6786], + device='cuda:1'), covar=tensor([0.0523, 0.0470, 0.0416, 0.0454, 0.0312, 0.0442, 0.0352, 0.0471], + device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0205, 0.0147, 0.0148, 0.0128, 0.0188, 0.0160, 0.0123], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:39:54,688 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=43083.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:40:08,149 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-20 23:40:08,542 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 +2023-03-20 23:40:08,623 INFO [train.py:901] (1/2) Epoch 16, batch 750, loss[loss=0.1777, simple_loss=0.2447, pruned_loss=0.05536, over 7134.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2361, pruned_loss=0.04349, over 1411933.71 frames. ], batch size: 98, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:40:08,645 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-20 23:40:12,690 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.251e+02 2.316e+02 2.793e+02 3.292e+02 6.692e+02, threshold=5.586e+02, percent-clipped=4.0 +2023-03-20 23:40:21,360 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3583, 3.0581, 3.0781, 2.9745, 3.4619, 3.0168, 3.0062, 3.2479], + device='cuda:1'), covar=tensor([0.1508, 0.0581, 0.1936, 0.1859, 0.0843, 0.1466, 0.2079, 0.1575], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0047, 0.0038, 0.0040, 0.0037, 0.0034, 0.0053, 0.0041], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:40:22,724 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-20 23:40:27,205 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-20 23:40:32,700 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-20 23:40:34,273 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-20 23:40:34,749 INFO [train.py:901] (1/2) Epoch 16, batch 800, loss[loss=0.1586, simple_loss=0.2356, pruned_loss=0.04081, over 7254.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2366, pruned_loss=0.04351, over 1419337.01 frames. ], batch size: 89, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:40:35,372 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43162.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:40:45,260 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-20 23:40:47,385 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2189, 1.3005, 1.2074, 1.2770, 1.3851, 1.1643, 1.2641, 0.8451], + device='cuda:1'), covar=tensor([0.0114, 0.0105, 0.0152, 0.0072, 0.0124, 0.0142, 0.0105, 0.0142], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0022, 0.0021, 0.0022, 0.0022, 0.0022, 0.0023, 0.0029], + device='cuda:1'), out_proj_covar=tensor([2.7868e-05, 2.4776e-05, 2.5537e-05, 2.5132e-05, 2.7118e-05, 2.4695e-05, + 2.6500e-05, 3.3700e-05], device='cuda:1') +2023-03-20 23:40:53,458 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3507, 1.4828, 1.3161, 1.2477, 1.5779, 1.3502, 1.4700, 1.0172], + device='cuda:1'), covar=tensor([0.0084, 0.0084, 0.0115, 0.0076, 0.0067, 0.0116, 0.0081, 0.0122], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0022, 0.0022, 0.0023, 0.0022, 0.0022, 0.0023, 0.0029], + device='cuda:1'), out_proj_covar=tensor([2.8013e-05, 2.4929e-05, 2.5637e-05, 2.5238e-05, 2.7154e-05, 2.4812e-05, + 2.6762e-05, 3.3730e-05], device='cuda:1') +2023-03-20 23:40:59,846 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=43210.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:41:00,280 INFO [train.py:901] (1/2) Epoch 16, batch 850, loss[loss=0.167, simple_loss=0.2464, pruned_loss=0.04382, over 7126.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2356, pruned_loss=0.04338, over 1420986.87 frames. ], batch size: 98, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:41:02,351 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43215.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:41:04,279 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.177e+02 2.571e+02 2.944e+02 4.753e+02, threshold=5.142e+02, percent-clipped=0.0 +2023-03-20 23:41:05,835 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-20 23:41:05,844 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-20 23:41:12,008 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-20 23:41:13,086 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43235.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:41:15,578 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-20 23:41:26,687 INFO [train.py:901] (1/2) Epoch 16, batch 900, loss[loss=0.157, simple_loss=0.2343, pruned_loss=0.03982, over 7281.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2358, pruned_loss=0.04349, over 1427152.56 frames. ], batch size: 68, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:41:49,021 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43305.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:41:52,504 INFO [train.py:901] (1/2) Epoch 16, batch 950, loss[loss=0.1445, simple_loss=0.2314, pruned_loss=0.02875, over 7265.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2367, pruned_loss=0.04388, over 1430141.29 frames. ], batch size: 57, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:41:52,515 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-20 23:41:56,554 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.071e+02 2.459e+02 3.042e+02 5.230e+02, threshold=4.919e+02, percent-clipped=1.0 +2023-03-20 23:42:14,278 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=43353.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:42:15,252 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-20 23:42:18,323 INFO [train.py:901] (1/2) Epoch 16, batch 1000, loss[loss=0.1365, simple_loss=0.1954, pruned_loss=0.03877, over 5933.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2371, pruned_loss=0.0439, over 1429171.14 frames. ], batch size: 25, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:42:36,578 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-20 23:42:44,396 INFO [train.py:901] (1/2) Epoch 16, batch 1050, loss[loss=0.179, simple_loss=0.2515, pruned_loss=0.05326, over 7239.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.237, pruned_loss=0.0439, over 1431624.75 frames. ], batch size: 45, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:42:48,406 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 2.268e+02 2.632e+02 3.225e+02 5.905e+02, threshold=5.263e+02, percent-clipped=2.0 +2023-03-20 23:42:58,732 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-20 23:43:03,239 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-20 23:43:10,294 INFO [train.py:901] (1/2) Epoch 16, batch 1100, loss[loss=0.1627, simple_loss=0.2376, pruned_loss=0.04395, over 7322.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2369, pruned_loss=0.04378, over 1432773.43 frames. ], batch size: 75, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:43:12,184 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-03-20 23:43:32,737 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-20 23:43:33,227 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 23:43:36,284 INFO [train.py:901] (1/2) Epoch 16, batch 1150, loss[loss=0.1653, simple_loss=0.2344, pruned_loss=0.04807, over 7271.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2362, pruned_loss=0.04365, over 1433716.21 frames. ], batch size: 52, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:43:38,536 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43514.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:43:39,053 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43515.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:43:39,103 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6457, 3.7083, 2.4767, 4.2133, 3.5663, 3.8143, 2.2955, 2.3336], + device='cuda:1'), covar=tensor([0.0275, 0.0571, 0.2435, 0.0341, 0.0646, 0.1015, 0.2589, 0.2030], + device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0238, 0.0304, 0.0242, 0.0246, 0.0254, 0.0265, 0.0286], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-20 23:43:40,853 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.302e+02 2.117e+02 2.537e+02 3.053e+02 5.220e+02, threshold=5.075e+02, percent-clipped=0.0 +2023-03-20 23:43:46,509 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-20 23:43:47,015 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-20 23:43:49,066 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43535.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:44:01,966 INFO [train.py:901] (1/2) Epoch 16, batch 1200, loss[loss=0.1884, simple_loss=0.2579, pruned_loss=0.05944, over 6726.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2366, pruned_loss=0.0439, over 1434857.41 frames. ], batch size: 106, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:44:03,076 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=43563.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:44:09,841 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43575.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:44:13,718 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=43583.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:44:18,690 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-20 23:44:28,672 INFO [train.py:901] (1/2) Epoch 16, batch 1250, loss[loss=0.1265, simple_loss=0.2102, pruned_loss=0.0214, over 7340.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2373, pruned_loss=0.04395, over 1437973.96 frames. ], batch size: 44, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:44:32,682 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.619e+02 2.248e+02 2.795e+02 3.324e+02 5.636e+02, threshold=5.591e+02, percent-clipped=2.0 +2023-03-20 23:44:37,885 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2991, 1.5023, 1.2545, 1.2963, 1.4823, 1.2913, 1.3810, 0.9219], + device='cuda:1'), covar=tensor([0.0118, 0.0108, 0.0142, 0.0088, 0.0084, 0.0079, 0.0115, 0.0116], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0022, 0.0022, 0.0023, 0.0023, 0.0022, 0.0024, 0.0029], + device='cuda:1'), out_proj_covar=tensor([2.8245e-05, 2.5702e-05, 2.5610e-05, 2.5557e-05, 2.7456e-05, 2.4366e-05, + 2.7381e-05, 3.3785e-05], device='cuda:1') +2023-03-20 23:44:42,925 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-20 23:44:46,530 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-20 23:44:46,812 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-03-20 23:44:47,969 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-20 23:44:50,107 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.2519, 4.8055, 4.6959, 5.3131, 5.1756, 5.2849, 4.7366, 4.8595], + device='cuda:1'), covar=tensor([0.0596, 0.2150, 0.1696, 0.0733, 0.0663, 0.0986, 0.0580, 0.1057], + device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0317, 0.0251, 0.0243, 0.0186, 0.0302, 0.0174, 0.0223], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:44:54,626 INFO [train.py:901] (1/2) Epoch 16, batch 1300, loss[loss=0.1611, simple_loss=0.2457, pruned_loss=0.03827, over 7130.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2373, pruned_loss=0.04377, over 1438647.86 frames. ], batch size: 98, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:45:02,447 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-03-20 23:45:11,976 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-20 23:45:14,426 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-20 23:45:17,879 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-20 23:45:20,337 INFO [train.py:901] (1/2) Epoch 16, batch 1350, loss[loss=0.158, simple_loss=0.2342, pruned_loss=0.04087, over 7288.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2375, pruned_loss=0.04407, over 1439989.27 frames. ], batch size: 68, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:45:24,366 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.464e+02 2.176e+02 2.636e+02 3.037e+02 1.340e+03, threshold=5.272e+02, percent-clipped=1.0 +2023-03-20 23:45:27,084 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-20 23:45:28,032 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-20 23:45:32,690 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 +2023-03-20 23:45:35,092 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9222, 3.6804, 4.0248, 3.8613, 4.1556, 3.9809, 3.9996, 3.9517], + device='cuda:1'), covar=tensor([0.0032, 0.0068, 0.0025, 0.0033, 0.0021, 0.0031, 0.0033, 0.0038], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0048, 0.0043, 0.0041, 0.0041, 0.0044, 0.0046, 0.0052], + device='cuda:1'), out_proj_covar=tensor([8.1424e-05, 1.2255e-04, 1.0695e-04, 9.1674e-05, 9.1877e-05, 1.0095e-04, + 1.1520e-04, 1.2108e-04], device='cuda:1') +2023-03-20 23:45:36,337 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-20 23:45:46,728 INFO [train.py:901] (1/2) Epoch 16, batch 1400, loss[loss=0.143, simple_loss=0.222, pruned_loss=0.032, over 7155.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2363, pruned_loss=0.04348, over 1440753.67 frames. ], batch size: 41, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:45:58,342 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 +2023-03-20 23:46:00,586 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43787.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:46:01,546 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-20 23:46:10,555 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3946, 4.3901, 4.1813, 3.7605, 3.9208, 2.6546, 1.8598, 4.5164], + device='cuda:1'), covar=tensor([0.0032, 0.0030, 0.0057, 0.0056, 0.0062, 0.0345, 0.0514, 0.0032], + device='cuda:1'), in_proj_covar=tensor([0.0081, 0.0068, 0.0091, 0.0076, 0.0099, 0.0114, 0.0115, 0.0081], + device='cuda:1'), out_proj_covar=tensor([1.0240e-04, 9.6980e-05, 1.1980e-04, 1.0387e-04, 1.2852e-04, 1.4960e-04, + 1.5134e-04, 1.0380e-04], device='cuda:1') +2023-03-20 23:46:12,340 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-20 23:46:12,960 INFO [train.py:901] (1/2) Epoch 16, batch 1450, loss[loss=0.1536, simple_loss=0.2308, pruned_loss=0.03818, over 7284.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2361, pruned_loss=0.04305, over 1440212.49 frames. ], batch size: 66, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:46:16,888 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.415e+02 2.136e+02 2.517e+02 2.991e+02 6.215e+02, threshold=5.033e+02, percent-clipped=2.0 +2023-03-20 23:46:20,004 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5214, 4.0725, 4.0556, 4.4364, 4.4816, 4.4726, 3.8937, 4.0712], + device='cuda:1'), covar=tensor([0.0728, 0.2479, 0.2147, 0.1027, 0.0644, 0.1062, 0.0729, 0.1140], + device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0311, 0.0250, 0.0245, 0.0183, 0.0297, 0.0174, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:46:25,560 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-20 23:46:32,241 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43848.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:46:35,774 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1496, 3.6157, 3.8267, 3.7910, 3.7098, 3.8279, 4.0467, 3.4440], + device='cuda:1'), covar=tensor([0.0110, 0.0156, 0.0126, 0.0142, 0.0375, 0.0095, 0.0134, 0.0176], + device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0076, 0.0076, 0.0064, 0.0132, 0.0085, 0.0080, 0.0081], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:46:39,412 INFO [train.py:901] (1/2) Epoch 16, batch 1500, loss[loss=0.169, simple_loss=0.2448, pruned_loss=0.04663, over 7323.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2362, pruned_loss=0.04293, over 1442521.61 frames. ], batch size: 54, lr: 1.03e-02, grad_scale: 8.0 +2023-03-20 23:46:42,479 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-20 23:46:44,096 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43870.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:46:49,110 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5457, 1.4260, 1.7158, 1.4771, 1.7432, 1.3082, 1.2497, 1.2237], + device='cuda:1'), covar=tensor([0.0410, 0.0454, 0.0179, 0.0139, 0.0508, 0.0381, 0.0356, 0.0344], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0024, 0.0022, 0.0021, 0.0021, 0.0020, 0.0023, 0.0025], + device='cuda:1'), out_proj_covar=tensor([6.0553e-05, 5.9890e-05, 5.4593e-05, 4.9908e-05, 5.5891e-05, 5.3171e-05, + 5.7464e-05, 6.2780e-05], device='cuda:1') +2023-03-20 23:47:02,844 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7797, 2.2886, 1.7198, 2.5980, 2.6667, 2.3695, 2.2172, 2.1498], + device='cuda:1'), covar=tensor([0.1847, 0.0631, 0.2834, 0.0545, 0.0124, 0.0061, 0.0141, 0.0212], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0228, 0.0266, 0.0256, 0.0142, 0.0133, 0.0161, 0.0185], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:47:03,855 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43909.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:47:04,717 INFO [train.py:901] (1/2) Epoch 16, batch 1550, loss[loss=0.1475, simple_loss=0.2363, pruned_loss=0.02941, over 7276.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2363, pruned_loss=0.04297, over 1444024.43 frames. ], batch size: 77, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:47:05,764 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-20 23:47:09,380 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 2.273e+02 2.649e+02 3.291e+02 5.530e+02, threshold=5.297e+02, percent-clipped=1.0 +2023-03-20 23:47:10,042 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7601, 1.7927, 2.1288, 1.7512, 2.1610, 1.6394, 1.7182, 1.4594], + device='cuda:1'), covar=tensor([0.0424, 0.0293, 0.0249, 0.0217, 0.0309, 0.0622, 0.0213, 0.0261], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0024, 0.0022, 0.0021, 0.0022, 0.0021, 0.0023, 0.0025], + device='cuda:1'), out_proj_covar=tensor([6.0844e-05, 5.9943e-05, 5.4625e-05, 5.0436e-05, 5.6437e-05, 5.3502e-05, + 5.7510e-05, 6.2764e-05], device='cuda:1') +2023-03-20 23:47:17,582 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5387, 3.0650, 3.2932, 3.2609, 2.7024, 2.7981, 3.4145, 2.8981], + device='cuda:1'), covar=tensor([0.0203, 0.0256, 0.0293, 0.0273, 0.0340, 0.0471, 0.0396, 0.0853], + device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0307, 0.0253, 0.0326, 0.0302, 0.0297, 0.0312, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-20 23:47:30,259 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43959.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:47:31,151 INFO [train.py:901] (1/2) Epoch 16, batch 1600, loss[loss=0.1523, simple_loss=0.2324, pruned_loss=0.03615, over 7335.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2369, pruned_loss=0.0436, over 1444227.58 frames. ], batch size: 75, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:47:35,819 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43970.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:47:36,168 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-20 23:47:37,145 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-20 23:47:39,680 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-20 23:47:50,705 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-20 23:47:55,097 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 +2023-03-20 23:47:59,312 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-20 23:48:00,779 INFO [train.py:901] (1/2) Epoch 16, batch 1650, loss[loss=0.1554, simple_loss=0.2263, pruned_loss=0.04221, over 7228.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.237, pruned_loss=0.04368, over 1442445.32 frames. ], batch size: 45, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:48:04,867 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.383e+02 2.044e+02 2.444e+02 3.043e+02 5.639e+02, threshold=4.888e+02, percent-clipped=2.0 +2023-03-20 23:48:05,551 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44020.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:48:07,499 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-20 23:48:11,671 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6134, 3.8784, 4.2247, 4.1226, 4.0076, 4.0663, 4.4317, 3.7932], + device='cuda:1'), covar=tensor([0.0118, 0.0155, 0.0108, 0.0123, 0.0382, 0.0115, 0.0126, 0.0182], + device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0075, 0.0077, 0.0063, 0.0133, 0.0087, 0.0082, 0.0082], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:48:24,785 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2304, 3.8679, 4.2367, 4.0789, 4.1324, 4.1585, 4.2209, 4.0365], + device='cuda:1'), covar=tensor([0.0026, 0.0067, 0.0027, 0.0033, 0.0031, 0.0031, 0.0027, 0.0047], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0048, 0.0043, 0.0040, 0.0041, 0.0044, 0.0045, 0.0052], + device='cuda:1'), out_proj_covar=tensor([7.9833e-05, 1.2198e-04, 1.0744e-04, 9.0369e-05, 9.2240e-05, 9.9759e-05, + 1.1366e-04, 1.1990e-04], device='cuda:1') +2023-03-20 23:48:25,195 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-20 23:48:26,658 INFO [train.py:901] (1/2) Epoch 16, batch 1700, loss[loss=0.1519, simple_loss=0.229, pruned_loss=0.03737, over 7311.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2364, pruned_loss=0.04353, over 1443346.10 frames. ], batch size: 80, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:48:29,271 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-20 23:48:35,929 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5578, 3.2247, 3.3672, 3.2372, 2.3616, 3.1324, 3.3203, 3.0161], + device='cuda:1'), covar=tensor([0.0286, 0.0324, 0.0242, 0.0323, 0.1052, 0.0334, 0.0461, 0.0420], + device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0077, 0.0077, 0.0064, 0.0133, 0.0087, 0.0083, 0.0082], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:48:39,406 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-20 23:48:52,748 INFO [train.py:901] (1/2) Epoch 16, batch 1750, loss[loss=0.1601, simple_loss=0.2329, pruned_loss=0.0436, over 7341.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2367, pruned_loss=0.04336, over 1443961.25 frames. ], batch size: 61, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:48:55,165 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-03-20 23:48:56,793 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.516e+02 2.154e+02 2.480e+02 3.011e+02 7.790e+02, threshold=4.959e+02, percent-clipped=2.0 +2023-03-20 23:49:05,039 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-20 23:49:05,543 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-20 23:49:09,678 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44143.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:49:17,325 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-03-20 23:49:18,531 INFO [train.py:901] (1/2) Epoch 16, batch 1800, loss[loss=0.1878, simple_loss=0.2615, pruned_loss=0.05706, over 7259.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2367, pruned_loss=0.04356, over 1445191.40 frames. ], batch size: 64, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:49:23,125 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44170.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:49:26,642 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-20 23:49:39,921 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-20 23:49:45,065 INFO [train.py:901] (1/2) Epoch 16, batch 1850, loss[loss=0.1762, simple_loss=0.2539, pruned_loss=0.04927, over 7275.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2368, pruned_loss=0.0435, over 1445432.01 frames. ], batch size: 57, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:49:48,574 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=44218.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:49:48,968 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.288e+02 2.195e+02 2.582e+02 3.217e+02 6.301e+02, threshold=5.165e+02, percent-clipped=5.0 +2023-03-20 23:49:50,998 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-20 23:50:08,095 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-20 23:50:10,095 INFO [train.py:901] (1/2) Epoch 16, batch 1900, loss[loss=0.1615, simple_loss=0.2316, pruned_loss=0.04576, over 7351.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2359, pruned_loss=0.04322, over 1444737.27 frames. ], batch size: 61, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:50:12,159 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44265.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:50:33,883 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-20 23:50:36,345 INFO [train.py:901] (1/2) Epoch 16, batch 1950, loss[loss=0.1651, simple_loss=0.2472, pruned_loss=0.04144, over 6785.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2356, pruned_loss=0.04291, over 1441698.83 frames. ], batch size: 107, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:50:38,457 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44315.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:50:40,373 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.314e+02 2.194e+02 2.559e+02 3.068e+02 6.748e+02, threshold=5.118e+02, percent-clipped=3.0 +2023-03-20 23:50:41,184 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-20 23:50:44,422 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-20 23:50:48,765 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-20 23:50:49,024 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-20 23:50:49,510 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-20 23:51:02,206 INFO [train.py:901] (1/2) Epoch 16, batch 2000, loss[loss=0.1602, simple_loss=0.2333, pruned_loss=0.04359, over 7319.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2368, pruned_loss=0.0435, over 1442242.23 frames. ], batch size: 83, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:51:05,731 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-20 23:51:05,837 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8473, 3.3618, 3.7956, 3.7885, 3.7339, 3.8035, 3.8849, 3.6432], + device='cuda:1'), covar=tensor([0.0025, 0.0083, 0.0034, 0.0037, 0.0032, 0.0036, 0.0039, 0.0051], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0048, 0.0044, 0.0040, 0.0041, 0.0043, 0.0045, 0.0052], + device='cuda:1'), out_proj_covar=tensor([7.8263e-05, 1.2121e-04, 1.0810e-04, 9.0346e-05, 9.1709e-05, 9.8411e-05, + 1.1233e-04, 1.1923e-04], device='cuda:1') +2023-03-20 23:51:16,897 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-20 23:51:21,713 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-03-20 23:51:24,910 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1544, 2.7712, 1.8617, 2.8511, 2.9749, 2.8731, 2.6199, 2.4560], + device='cuda:1'), covar=tensor([0.1448, 0.0601, 0.2915, 0.0617, 0.0093, 0.0074, 0.0217, 0.0313], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0228, 0.0267, 0.0261, 0.0144, 0.0132, 0.0163, 0.0187], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:51:25,777 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-20 23:51:28,248 INFO [train.py:901] (1/2) Epoch 16, batch 2050, loss[loss=0.1444, simple_loss=0.2237, pruned_loss=0.03251, over 7328.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2362, pruned_loss=0.04306, over 1441026.73 frames. ], batch size: 61, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:51:32,327 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.233e+02 2.247e+02 2.734e+02 3.151e+02 5.434e+02, threshold=5.468e+02, percent-clipped=2.0 +2023-03-20 23:51:45,052 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44443.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:51:54,001 INFO [train.py:901] (1/2) Epoch 16, batch 2100, loss[loss=0.1416, simple_loss=0.2212, pruned_loss=0.03098, over 7328.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2356, pruned_loss=0.04285, over 1440708.79 frames. ], batch size: 44, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:52:00,447 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-20 23:52:04,554 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-20 23:52:09,596 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=44491.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:52:19,293 INFO [train.py:901] (1/2) Epoch 16, batch 2150, loss[loss=0.1728, simple_loss=0.2419, pruned_loss=0.0518, over 7273.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2361, pruned_loss=0.043, over 1442594.03 frames. ], batch size: 77, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:52:23,162 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.579e+02 2.371e+02 2.688e+02 3.151e+02 5.325e+02, threshold=5.377e+02, percent-clipped=0.0 +2023-03-20 23:52:27,906 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44527.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:52:39,487 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44550.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:52:44,913 INFO [train.py:901] (1/2) Epoch 16, batch 2200, loss[loss=0.1723, simple_loss=0.2445, pruned_loss=0.05009, over 7305.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2361, pruned_loss=0.04315, over 1442231.46 frames. ], batch size: 59, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:52:47,065 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44565.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:52:49,002 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-20 23:52:50,931 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 +2023-03-20 23:52:58,286 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9821, 4.4381, 4.4870, 4.4399, 4.4731, 4.0675, 4.5331, 4.4136], + device='cuda:1'), covar=tensor([0.0559, 0.0498, 0.0479, 0.0533, 0.0321, 0.0423, 0.0388, 0.0461], + device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0215, 0.0155, 0.0157, 0.0132, 0.0196, 0.0166, 0.0128], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:52:59,355 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44588.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:53:11,849 INFO [train.py:901] (1/2) Epoch 16, batch 2250, loss[loss=0.1619, simple_loss=0.2348, pruned_loss=0.0445, over 7324.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2364, pruned_loss=0.04286, over 1443836.65 frames. ], batch size: 59, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:53:12,005 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44611.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:53:12,948 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=44613.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:53:14,029 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44615.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:53:15,906 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.027e+02 2.444e+02 2.860e+02 7.319e+02, threshold=4.889e+02, percent-clipped=1.0 +2023-03-20 23:53:22,505 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-20 23:53:22,921 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-20 23:53:36,689 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-20 23:53:37,646 INFO [train.py:901] (1/2) Epoch 16, batch 2300, loss[loss=0.1791, simple_loss=0.2506, pruned_loss=0.0538, over 7325.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2364, pruned_loss=0.04286, over 1444391.68 frames. ], batch size: 61, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:53:38,735 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=44663.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:53:41,361 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9926, 2.4917, 2.9339, 2.6804, 3.0972, 2.6512, 2.4458, 2.9834], + device='cuda:1'), covar=tensor([0.1606, 0.0687, 0.1321, 0.2462, 0.0821, 0.1359, 0.2439, 0.1295], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0048, 0.0038, 0.0039, 0.0038, 0.0035, 0.0053, 0.0041], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 23:54:03,687 INFO [train.py:901] (1/2) Epoch 16, batch 2350, loss[loss=0.1667, simple_loss=0.2383, pruned_loss=0.04756, over 7263.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2365, pruned_loss=0.04305, over 1443978.31 frames. ], batch size: 66, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:54:03,968 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-20 23:54:07,744 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 2.179e+02 2.590e+02 3.054e+02 7.266e+02, threshold=5.180e+02, percent-clipped=2.0 +2023-03-20 23:54:24,454 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-20 23:54:25,572 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9346, 3.2232, 2.6171, 2.7903, 2.9231, 2.4959, 3.0859, 2.9356], + device='cuda:1'), covar=tensor([0.0580, 0.1221, 0.1029, 0.1492, 0.1424, 0.0847, 0.1495, 0.1355], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0043, 0.0049, 0.0043, 0.0042, 0.0045, 0.0044, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:54:29,406 INFO [train.py:901] (1/2) Epoch 16, batch 2400, loss[loss=0.1677, simple_loss=0.2473, pruned_loss=0.04405, over 7316.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.236, pruned_loss=0.04274, over 1444181.56 frames. ], batch size: 59, lr: 1.02e-02, grad_scale: 8.0 +2023-03-20 23:54:30,436 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-20 23:54:42,171 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-20 23:54:44,707 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-20 23:54:55,557 INFO [train.py:901] (1/2) Epoch 16, batch 2450, loss[loss=0.1305, simple_loss=0.2017, pruned_loss=0.02968, over 7170.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.236, pruned_loss=0.04274, over 1445887.27 frames. ], batch size: 39, lr: 1.01e-02, grad_scale: 8.0 +2023-03-20 23:54:59,657 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.397e+02 2.079e+02 2.447e+02 2.964e+02 4.807e+02, threshold=4.894e+02, percent-clipped=0.0 +2023-03-20 23:55:10,786 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-20 23:55:21,250 INFO [train.py:901] (1/2) Epoch 16, batch 2500, loss[loss=0.1225, simple_loss=0.1679, pruned_loss=0.03852, over 6312.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2359, pruned_loss=0.04296, over 1444013.01 frames. ], batch size: 27, lr: 1.01e-02, grad_scale: 16.0 +2023-03-20 23:55:21,839 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4776, 5.0330, 5.1468, 5.0331, 4.8817, 4.6317, 5.1731, 4.8832], + device='cuda:1'), covar=tensor([0.0429, 0.0384, 0.0359, 0.0455, 0.0313, 0.0295, 0.0300, 0.0479], + device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0211, 0.0154, 0.0156, 0.0129, 0.0192, 0.0165, 0.0126], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:55:32,925 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44883.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:55:35,762 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-20 23:55:38,380 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7411, 1.9985, 1.6032, 2.6291, 2.6167, 2.7255, 2.3026, 2.0642], + device='cuda:1'), covar=tensor([0.1712, 0.0844, 0.3151, 0.0547, 0.0130, 0.0086, 0.0250, 0.0273], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0227, 0.0267, 0.0256, 0.0145, 0.0134, 0.0164, 0.0186], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:55:44,404 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44906.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:55:46,772 INFO [train.py:901] (1/2) Epoch 16, batch 2550, loss[loss=0.1708, simple_loss=0.2452, pruned_loss=0.04819, over 7306.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2363, pruned_loss=0.04343, over 1441577.53 frames. ], batch size: 59, lr: 1.01e-02, grad_scale: 16.0 +2023-03-20 23:55:50,067 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4925, 1.3103, 1.6906, 1.9911, 1.7994, 2.0986, 1.8443, 1.9359], + device='cuda:1'), covar=tensor([0.1660, 0.2086, 0.1367, 0.1233, 0.2002, 0.0992, 0.1962, 0.6265], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0050, 0.0037, 0.0036, 0.0041, 0.0043, 0.0056, 0.0043], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-20 23:55:51,395 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.227e+02 2.271e+02 2.744e+02 3.382e+02 5.975e+02, threshold=5.489e+02, percent-clipped=3.0 +2023-03-20 23:55:54,661 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-03-20 23:56:12,442 INFO [train.py:901] (1/2) Epoch 16, batch 2600, loss[loss=0.1463, simple_loss=0.2289, pruned_loss=0.03184, over 7327.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2354, pruned_loss=0.04291, over 1441099.98 frames. ], batch size: 75, lr: 1.01e-02, grad_scale: 16.0 +2023-03-20 23:56:19,762 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.82 vs. limit=5.0 +2023-03-20 23:56:21,031 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0453, 2.3197, 1.6664, 2.5798, 2.8634, 2.8256, 2.5504, 2.6010], + device='cuda:1'), covar=tensor([0.1642, 0.0680, 0.3234, 0.0369, 0.0095, 0.0056, 0.0194, 0.0284], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0228, 0.0268, 0.0257, 0.0145, 0.0135, 0.0164, 0.0186], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:56:38,043 INFO [train.py:901] (1/2) Epoch 16, batch 2650, loss[loss=0.1951, simple_loss=0.266, pruned_loss=0.06207, over 7125.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2357, pruned_loss=0.04281, over 1441521.15 frames. ], batch size: 98, lr: 1.01e-02, grad_scale: 16.0 +2023-03-20 23:56:42,040 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.330e+02 2.314e+02 2.696e+02 3.336e+02 6.334e+02, threshold=5.393e+02, percent-clipped=2.0 +2023-03-20 23:57:00,351 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45056.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:57:02,757 INFO [train.py:901] (1/2) Epoch 16, batch 2700, loss[loss=0.1802, simple_loss=0.2559, pruned_loss=0.05226, over 7262.00 frames. ], tot_loss[loss=0.161, simple_loss=0.236, pruned_loss=0.04304, over 1441816.69 frames. ], batch size: 55, lr: 1.01e-02, grad_scale: 16.0 +2023-03-20 23:57:11,056 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 +2023-03-20 23:57:12,721 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45081.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:57:27,939 INFO [train.py:901] (1/2) Epoch 16, batch 2750, loss[loss=0.1567, simple_loss=0.2283, pruned_loss=0.04259, over 7274.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2358, pruned_loss=0.04316, over 1440451.61 frames. ], batch size: 47, lr: 1.01e-02, grad_scale: 16.0 +2023-03-20 23:57:31,027 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45117.0, num_to_drop=1, layers_to_drop={2} +2023-03-20 23:57:31,799 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.386e+02 1.985e+02 2.477e+02 3.081e+02 5.070e+02, threshold=4.954e+02, percent-clipped=0.0 +2023-03-20 23:57:32,904 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0829, 3.8384, 3.7218, 3.9208, 3.0702, 3.7515, 3.7957, 3.5827], + device='cuda:1'), covar=tensor([0.0195, 0.0161, 0.0205, 0.0169, 0.0648, 0.0156, 0.0292, 0.0184], + device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0077, 0.0078, 0.0066, 0.0134, 0.0088, 0.0083, 0.0083], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:57:43,278 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45142.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:57:48,409 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 +2023-03-20 23:57:49,290 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.14 vs. limit=2.0 +2023-03-20 23:57:52,358 INFO [train.py:901] (1/2) Epoch 16, batch 2800, loss[loss=0.1756, simple_loss=0.2572, pruned_loss=0.04698, over 7238.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2369, pruned_loss=0.04398, over 1441388.22 frames. ], batch size: 55, lr: 1.01e-02, grad_scale: 8.0 +2023-03-20 23:58:03,081 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45183.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:58:17,864 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-20 23:58:17,916 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-20 23:58:17,940 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-20 23:58:17,956 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-20 23:58:18,447 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-20 23:58:18,461 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-20 23:58:18,478 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-20 23:58:18,574 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-20 23:58:18,647 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-20 23:58:18,662 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-20 23:58:18,678 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-20 23:58:18,843 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-20 23:58:18,849 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-20 23:58:18,856 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-20 23:58:18,993 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-20 23:58:18,996 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-20 23:58:19,390 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-20 23:58:25,824 INFO [train.py:901] (1/2) Epoch 17, batch 0, loss[loss=0.1567, simple_loss=0.2337, pruned_loss=0.03982, over 7329.00 frames. ], tot_loss[loss=0.1567, simple_loss=0.2337, pruned_loss=0.03982, over 7329.00 frames. ], batch size: 61, lr: 9.83e-03, grad_scale: 8.0 +2023-03-20 23:58:25,824 INFO [train.py:926] (1/2) Computing validation loss +2023-03-20 23:58:35,427 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7503, 2.3309, 2.7699, 2.8519, 2.9923, 2.5805, 2.3842, 3.1338], + device='cuda:1'), covar=tensor([0.2103, 0.0706, 0.2272, 0.1865, 0.1243, 0.1431, 0.2449, 0.1105], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0049, 0.0038, 0.0039, 0.0039, 0.0036, 0.0054, 0.0041], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-20 23:58:52,331 INFO [train.py:935] (1/2) Epoch 17, validation: loss=0.1675, simple_loss=0.2548, pruned_loss=0.04006, over 1622729.00 frames. +2023-03-20 23:58:52,332 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-20 23:58:58,905 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-20 23:59:03,994 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7036, 4.3266, 4.2403, 4.7049, 4.6540, 4.7007, 3.8710, 4.2242], + device='cuda:1'), covar=tensor([0.0796, 0.2716, 0.2428, 0.1149, 0.0820, 0.1240, 0.0908, 0.1177], + device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0322, 0.0251, 0.0250, 0.0188, 0.0309, 0.0180, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:59:04,029 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45206.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:59:09,922 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-20 23:59:10,865 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.483e+02 2.268e+02 2.699e+02 3.291e+02 1.082e+03, threshold=5.398e+02, percent-clipped=4.0 +2023-03-20 23:59:16,366 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-20 23:59:16,420 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=45231.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:59:18,368 INFO [train.py:901] (1/2) Epoch 17, batch 50, loss[loss=0.197, simple_loss=0.2659, pruned_loss=0.06405, over 6703.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2378, pruned_loss=0.04402, over 326316.57 frames. ], batch size: 107, lr: 9.82e-03, grad_scale: 8.0 +2023-03-20 23:59:18,879 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-20 23:59:21,845 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-20 23:59:23,978 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7995, 3.1683, 3.5946, 3.5285, 3.6716, 3.8228, 3.5456, 3.5606], + device='cuda:1'), covar=tensor([0.0022, 0.0081, 0.0034, 0.0043, 0.0033, 0.0022, 0.0047, 0.0045], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0049, 0.0044, 0.0042, 0.0042, 0.0044, 0.0044, 0.0053], + device='cuda:1'), out_proj_covar=tensor([7.9814e-05, 1.2302e-04, 1.0804e-04, 9.3631e-05, 9.5031e-05, 9.7633e-05, + 1.1015e-04, 1.2273e-04], device='cuda:1') +2023-03-20 23:59:27,937 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=45254.0, num_to_drop=0, layers_to_drop=set() +2023-03-20 23:59:27,968 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8693, 4.4082, 4.4871, 4.3904, 4.4517, 4.0733, 4.5139, 4.3378], + device='cuda:1'), covar=tensor([0.0589, 0.0495, 0.0426, 0.0480, 0.0340, 0.0373, 0.0359, 0.0555], + device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0208, 0.0154, 0.0151, 0.0128, 0.0189, 0.0161, 0.0124], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-20 23:59:37,984 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-20 23:59:38,493 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-20 23:59:43,999 INFO [train.py:901] (1/2) Epoch 17, batch 100, loss[loss=0.1497, simple_loss=0.231, pruned_loss=0.03422, over 7256.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2374, pruned_loss=0.04273, over 575334.21 frames. ], batch size: 64, lr: 9.82e-03, grad_scale: 8.0 +2023-03-20 23:59:52,304 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3462, 3.9641, 3.9691, 4.0801, 3.8231, 4.1306, 4.2517, 3.7592], + device='cuda:1'), covar=tensor([0.0148, 0.0140, 0.0173, 0.0124, 0.0439, 0.0094, 0.0154, 0.0151], + device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0077, 0.0079, 0.0067, 0.0136, 0.0088, 0.0084, 0.0084], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:00:02,109 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.391e+02 2.128e+02 2.527e+02 2.960e+02 4.350e+02, threshold=5.054e+02, percent-clipped=0.0 +2023-03-21 00:00:07,825 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8865, 3.5579, 3.6081, 3.6622, 3.5212, 3.5786, 3.7574, 3.3693], + device='cuda:1'), covar=tensor([0.0132, 0.0182, 0.0180, 0.0147, 0.0370, 0.0135, 0.0172, 0.0168], + device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0076, 0.0078, 0.0067, 0.0134, 0.0088, 0.0084, 0.0083], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:00:09,803 INFO [train.py:901] (1/2) Epoch 17, batch 150, loss[loss=0.1573, simple_loss=0.2272, pruned_loss=0.04374, over 7253.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.237, pruned_loss=0.04269, over 770309.58 frames. ], batch size: 55, lr: 9.81e-03, grad_scale: 8.0 +2023-03-21 00:00:10,970 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4428, 2.7147, 2.4004, 2.6320, 2.4981, 2.3109, 2.6630, 2.5329], + device='cuda:1'), covar=tensor([0.0919, 0.0918, 0.1079, 0.0862, 0.0973, 0.0827, 0.1355, 0.0878], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0043, 0.0050, 0.0044, 0.0042, 0.0044, 0.0045, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:00:16,609 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45348.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 00:00:22,314 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8931, 3.9029, 3.2680, 3.3107, 3.0933, 2.3758, 1.9387, 3.8845], + device='cuda:1'), covar=tensor([0.0035, 0.0043, 0.0088, 0.0056, 0.0104, 0.0388, 0.0469, 0.0037], + device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0070, 0.0091, 0.0076, 0.0099, 0.0115, 0.0115, 0.0082], + device='cuda:1'), out_proj_covar=tensor([1.0419e-04, 9.8419e-05, 1.1875e-04, 1.0331e-04, 1.2715e-04, 1.5138e-04, + 1.5161e-04, 1.0553e-04], device='cuda:1') +2023-03-21 00:00:36,392 INFO [train.py:901] (1/2) Epoch 17, batch 200, loss[loss=0.1304, simple_loss=0.2028, pruned_loss=0.02901, over 7146.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2359, pruned_loss=0.04222, over 919840.87 frames. ], batch size: 41, lr: 9.80e-03, grad_scale: 8.0 +2023-03-21 00:00:38,974 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 00:00:44,073 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 00:00:49,073 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45409.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 00:00:50,440 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 00:00:50,499 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45412.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 00:00:54,311 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.033e+02 2.393e+02 3.029e+02 5.248e+02, threshold=4.786e+02, percent-clipped=2.0 +2023-03-21 00:01:01,785 INFO [train.py:901] (1/2) Epoch 17, batch 250, loss[loss=0.2143, simple_loss=0.2814, pruned_loss=0.07358, over 6755.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2359, pruned_loss=0.0426, over 1033461.48 frames. ], batch size: 106, lr: 9.80e-03, grad_scale: 8.0 +2023-03-21 00:01:02,826 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 00:01:02,899 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45437.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:01:24,087 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 00:01:28,136 INFO [train.py:901] (1/2) Epoch 17, batch 300, loss[loss=0.1305, simple_loss=0.1928, pruned_loss=0.03412, over 6011.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2344, pruned_loss=0.04205, over 1120544.45 frames. ], batch size: 26, lr: 9.79e-03, grad_scale: 8.0 +2023-03-21 00:01:32,771 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 00:01:45,535 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.226e+02 2.225e+02 2.796e+02 3.288e+02 5.407e+02, threshold=5.591e+02, percent-clipped=4.0 +2023-03-21 00:01:53,617 INFO [train.py:901] (1/2) Epoch 17, batch 350, loss[loss=0.1582, simple_loss=0.2356, pruned_loss=0.04038, over 7276.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2355, pruned_loss=0.04269, over 1191604.41 frames. ], batch size: 70, lr: 9.79e-03, grad_scale: 8.0 +2023-03-21 00:02:06,379 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 00:02:08,297 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-21 00:02:09,846 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 +2023-03-21 00:02:19,555 INFO [train.py:901] (1/2) Epoch 17, batch 400, loss[loss=0.1639, simple_loss=0.2396, pruned_loss=0.04407, over 7286.00 frames. ], tot_loss[loss=0.1595, simple_loss=0.2347, pruned_loss=0.04217, over 1248570.01 frames. ], batch size: 52, lr: 9.78e-03, grad_scale: 8.0 +2023-03-21 00:02:26,391 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.49 vs. limit=5.0 +2023-03-21 00:02:27,620 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0402, 2.4371, 1.6753, 2.8161, 2.4846, 2.7263, 2.3249, 2.1200], + device='cuda:1'), covar=tensor([0.1800, 0.0771, 0.3240, 0.0461, 0.0104, 0.0078, 0.0198, 0.0202], + device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0236, 0.0270, 0.0267, 0.0148, 0.0142, 0.0170, 0.0190], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:02:38,158 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 2.037e+02 2.425e+02 2.810e+02 4.522e+02, threshold=4.850e+02, percent-clipped=0.0 +2023-03-21 00:02:42,074 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 +2023-03-21 00:02:45,676 INFO [train.py:901] (1/2) Epoch 17, batch 450, loss[loss=0.1635, simple_loss=0.2341, pruned_loss=0.04648, over 7264.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2351, pruned_loss=0.04236, over 1290799.75 frames. ], batch size: 47, lr: 9.78e-03, grad_scale: 8.0 +2023-03-21 00:02:49,815 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 00:02:50,312 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 00:02:53,122 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 +2023-03-21 00:03:11,149 INFO [train.py:901] (1/2) Epoch 17, batch 500, loss[loss=0.1501, simple_loss=0.2276, pruned_loss=0.03627, over 7278.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2355, pruned_loss=0.04233, over 1323935.04 frames. ], batch size: 52, lr: 9.77e-03, grad_scale: 8.0 +2023-03-21 00:03:20,489 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45702.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:03:21,433 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45704.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 00:03:22,854 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 00:03:24,378 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 00:03:24,910 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 00:03:25,556 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45712.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 00:03:27,469 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 00:03:29,417 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 2.314e+02 2.733e+02 3.177e+02 5.495e+02, threshold=5.467e+02, percent-clipped=2.0 +2023-03-21 00:03:31,622 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 00:03:34,366 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4113, 2.5658, 2.3130, 2.5906, 2.4641, 2.3376, 2.5591, 2.4832], + device='cuda:1'), covar=tensor([0.0793, 0.0614, 0.1325, 0.0863, 0.1276, 0.0786, 0.1560, 0.1166], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0045, 0.0050, 0.0044, 0.0043, 0.0044, 0.0045, 0.0043], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:03:37,733 INFO [train.py:901] (1/2) Epoch 17, batch 550, loss[loss=0.1579, simple_loss=0.2323, pruned_loss=0.04182, over 7221.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2354, pruned_loss=0.04243, over 1348632.72 frames. ], batch size: 93, lr: 9.77e-03, grad_scale: 8.0 +2023-03-21 00:03:38,899 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45737.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:03:42,343 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 00:03:50,355 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=45760.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:03:51,230 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 00:03:51,895 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45763.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:03:54,180 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 00:03:57,239 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45774.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:04:01,607 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 00:04:02,582 INFO [train.py:901] (1/2) Epoch 17, batch 600, loss[loss=0.1543, simple_loss=0.2404, pruned_loss=0.03416, over 7289.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2353, pruned_loss=0.04265, over 1367413.05 frames. ], batch size: 68, lr: 9.76e-03, grad_scale: 8.0 +2023-03-21 00:04:02,664 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=45785.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:04:10,515 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 +2023-03-21 00:04:19,239 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 00:04:21,687 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.428e+02 1.996e+02 2.450e+02 2.921e+02 7.497e+02, threshold=4.900e+02, percent-clipped=1.0 +2023-03-21 00:04:24,900 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6136, 1.8742, 2.1457, 2.0369, 1.9496, 1.7650, 1.7555, 1.4027], + device='cuda:1'), covar=tensor([0.0573, 0.0273, 0.0097, 0.0138, 0.0518, 0.0371, 0.0215, 0.0277], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0024, 0.0022, 0.0021, 0.0022, 0.0020, 0.0024, 0.0025], + device='cuda:1'), out_proj_covar=tensor([6.0889e-05, 6.0378e-05, 5.4907e-05, 5.0803e-05, 5.7340e-05, 5.3360e-05, + 5.8849e-05, 6.2920e-05], device='cuda:1') +2023-03-21 00:04:28,286 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 00:04:29,295 INFO [train.py:901] (1/2) Epoch 17, batch 650, loss[loss=0.1688, simple_loss=0.2409, pruned_loss=0.04836, over 7242.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2349, pruned_loss=0.04267, over 1383728.47 frames. ], batch size: 55, lr: 9.76e-03, grad_scale: 8.0 +2023-03-21 00:04:29,464 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45835.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:04:41,695 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45859.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:04:43,186 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3661, 3.2415, 2.0910, 3.9098, 2.3820, 2.9595, 1.7782, 2.0937], + device='cuda:1'), covar=tensor([0.0245, 0.0458, 0.1896, 0.0362, 0.0275, 0.0326, 0.2685, 0.1522], + device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0233, 0.0297, 0.0243, 0.0251, 0.0246, 0.0259, 0.0281], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 00:04:46,496 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 00:04:55,319 INFO [train.py:901] (1/2) Epoch 17, batch 700, loss[loss=0.1618, simple_loss=0.2429, pruned_loss=0.04035, over 7299.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2349, pruned_loss=0.04265, over 1396251.61 frames. ], batch size: 68, lr: 9.75e-03, grad_scale: 8.0 +2023-03-21 00:04:55,340 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 00:05:13,503 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.307e+02 2.037e+02 2.465e+02 3.045e+02 9.193e+02, threshold=4.929e+02, percent-clipped=2.0 +2023-03-21 00:05:13,653 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45920.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:05:19,561 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 00:05:19,579 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 00:05:21,059 INFO [train.py:901] (1/2) Epoch 17, batch 750, loss[loss=0.1774, simple_loss=0.2508, pruned_loss=0.05204, over 7231.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2345, pruned_loss=0.04212, over 1407632.78 frames. ], batch size: 89, lr: 9.75e-03, grad_scale: 8.0 +2023-03-21 00:05:33,144 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 00:05:37,793 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 00:05:43,755 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-03-21 00:05:44,259 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.80 vs. limit=5.0 +2023-03-21 00:05:44,481 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 00:05:46,023 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 00:05:47,524 INFO [train.py:901] (1/2) Epoch 17, batch 800, loss[loss=0.1595, simple_loss=0.2375, pruned_loss=0.04075, over 7264.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2359, pruned_loss=0.04274, over 1416769.13 frames. ], batch size: 70, lr: 9.74e-03, grad_scale: 8.0 +2023-03-21 00:05:56,074 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1955, 2.7275, 3.2401, 2.9739, 3.2373, 2.9478, 2.5162, 3.1664], + device='cuda:1'), covar=tensor([0.1403, 0.0632, 0.1059, 0.2086, 0.0938, 0.1013, 0.2585, 0.1512], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0048, 0.0038, 0.0039, 0.0037, 0.0035, 0.0053, 0.0041], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 00:05:57,461 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 00:05:57,566 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46004.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 00:06:05,490 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.317e+02 2.247e+02 2.680e+02 3.304e+02 7.121e+02, threshold=5.361e+02, percent-clipped=5.0 +2023-03-21 00:06:12,887 INFO [train.py:901] (1/2) Epoch 17, batch 850, loss[loss=0.1534, simple_loss=0.2267, pruned_loss=0.04009, over 7359.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2353, pruned_loss=0.04261, over 1419259.41 frames. ], batch size: 73, lr: 9.74e-03, grad_scale: 8.0 +2023-03-21 00:06:15,988 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 00:06:15,993 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 00:06:21,598 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 00:06:22,260 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=46052.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 00:06:25,284 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46058.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:06:25,687 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 00:06:32,522 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4657, 5.0590, 5.1592, 5.0843, 4.8634, 4.5832, 5.1296, 4.8932], + device='cuda:1'), covar=tensor([0.0507, 0.0373, 0.0280, 0.0365, 0.0307, 0.0324, 0.0293, 0.0530], + device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0208, 0.0152, 0.0153, 0.0126, 0.0191, 0.0163, 0.0123], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:06:39,598 INFO [train.py:901] (1/2) Epoch 17, batch 900, loss[loss=0.1563, simple_loss=0.2331, pruned_loss=0.03973, over 7261.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2349, pruned_loss=0.0423, over 1424533.49 frames. ], batch size: 89, lr: 9.73e-03, grad_scale: 8.0 +2023-03-21 00:06:42,621 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-03-21 00:06:48,628 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.60 vs. limit=5.0 +2023-03-21 00:06:57,328 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.352e+02 2.073e+02 2.414e+02 2.873e+02 6.303e+02, threshold=4.829e+02, percent-clipped=2.0 +2023-03-21 00:07:02,453 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46130.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:07:03,914 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 00:07:04,928 INFO [train.py:901] (1/2) Epoch 17, batch 950, loss[loss=0.1503, simple_loss=0.2246, pruned_loss=0.03802, over 7338.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2344, pruned_loss=0.04224, over 1425798.82 frames. ], batch size: 44, lr: 9.73e-03, grad_scale: 8.0 +2023-03-21 00:07:19,739 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9509, 3.3514, 3.9201, 3.9383, 3.9164, 4.0611, 3.9320, 3.7774], + device='cuda:1'), covar=tensor([0.0027, 0.0090, 0.0033, 0.0028, 0.0030, 0.0022, 0.0034, 0.0044], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0051, 0.0044, 0.0042, 0.0042, 0.0045, 0.0044, 0.0054], + device='cuda:1'), out_proj_covar=tensor([7.9790e-05, 1.2793e-04, 1.0670e-04, 9.3934e-05, 9.3427e-05, 9.7569e-05, + 1.0831e-04, 1.2507e-04], device='cuda:1') +2023-03-21 00:07:28,283 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 00:07:31,210 INFO [train.py:901] (1/2) Epoch 17, batch 1000, loss[loss=0.1565, simple_loss=0.2317, pruned_loss=0.04061, over 7356.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2343, pruned_loss=0.0423, over 1429184.01 frames. ], batch size: 63, lr: 9.72e-03, grad_scale: 8.0 +2023-03-21 00:07:46,769 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46215.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:07:49,125 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.429e+02 2.224e+02 2.568e+02 3.031e+02 6.347e+02, threshold=5.135e+02, percent-clipped=2.0 +2023-03-21 00:07:49,652 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 00:07:57,328 INFO [train.py:901] (1/2) Epoch 17, batch 1050, loss[loss=0.1733, simple_loss=0.2507, pruned_loss=0.04793, over 7288.00 frames. ], tot_loss[loss=0.1591, simple_loss=0.2338, pruned_loss=0.04223, over 1430257.99 frames. ], batch size: 70, lr: 9.71e-03, grad_scale: 8.0 +2023-03-21 00:08:02,531 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2415, 1.3836, 1.3157, 1.2667, 1.3451, 1.1973, 1.3575, 1.0308], + device='cuda:1'), covar=tensor([0.0109, 0.0112, 0.0205, 0.0085, 0.0099, 0.0107, 0.0120, 0.0113], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0022, 0.0021, 0.0022, 0.0023, 0.0022, 0.0024, 0.0029], + device='cuda:1'), out_proj_covar=tensor([2.8296e-05, 2.4846e-05, 2.5365e-05, 2.5202e-05, 2.6866e-05, 2.4490e-05, + 2.6980e-05, 3.4294e-05], device='cuda:1') +2023-03-21 00:08:12,255 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8560, 2.7292, 3.0357, 2.6339, 2.2857, 2.3497, 2.8896, 2.1776], + device='cuda:1'), covar=tensor([0.0324, 0.0302, 0.0302, 0.0336, 0.0378, 0.0569, 0.0398, 0.1155], + device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0316, 0.0254, 0.0340, 0.0306, 0.0306, 0.0321, 0.0294], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 00:08:12,579 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 00:08:16,633 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 00:08:18,746 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46276.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:08:23,171 INFO [train.py:901] (1/2) Epoch 17, batch 1100, loss[loss=0.158, simple_loss=0.2393, pruned_loss=0.03835, over 7145.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.2335, pruned_loss=0.04176, over 1432289.83 frames. ], batch size: 98, lr: 9.71e-03, grad_scale: 8.0 +2023-03-21 00:08:41,469 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 2.294e+02 2.723e+02 3.256e+02 7.179e+02, threshold=5.446e+02, percent-clipped=3.0 +2023-03-21 00:08:45,012 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 00:08:45,538 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 00:08:49,646 INFO [train.py:901] (1/2) Epoch 17, batch 1150, loss[loss=0.1761, simple_loss=0.252, pruned_loss=0.05005, over 7221.00 frames. ], tot_loss[loss=0.1591, simple_loss=0.2342, pruned_loss=0.04197, over 1435993.60 frames. ], batch size: 93, lr: 9.70e-03, grad_scale: 8.0 +2023-03-21 00:08:50,841 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46337.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:08:57,945 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 00:08:58,449 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 00:09:01,551 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46358.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:09:08,617 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1206, 3.9100, 3.8914, 3.8861, 3.1831, 3.8797, 4.0717, 3.7152], + device='cuda:1'), covar=tensor([0.0197, 0.0190, 0.0164, 0.0191, 0.0623, 0.0141, 0.0218, 0.0200], + device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0075, 0.0076, 0.0067, 0.0133, 0.0087, 0.0082, 0.0083], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:09:15,022 INFO [train.py:901] (1/2) Epoch 17, batch 1200, loss[loss=0.1701, simple_loss=0.2512, pruned_loss=0.0445, over 7224.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2347, pruned_loss=0.04203, over 1437286.47 frames. ], batch size: 93, lr: 9.70e-03, grad_scale: 8.0 +2023-03-21 00:09:24,652 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1432, 2.8275, 3.2890, 3.0601, 3.4316, 3.0835, 2.8119, 3.2790], + device='cuda:1'), covar=tensor([0.1571, 0.0755, 0.1015, 0.2315, 0.0676, 0.1174, 0.1655, 0.1332], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0047, 0.0037, 0.0039, 0.0036, 0.0034, 0.0052, 0.0040], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:09:25,142 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8663, 3.8880, 3.3573, 3.2841, 3.0001, 2.4111, 1.7990, 3.8403], + device='cuda:1'), covar=tensor([0.0032, 0.0041, 0.0079, 0.0070, 0.0104, 0.0382, 0.0493, 0.0038], + device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0069, 0.0089, 0.0078, 0.0098, 0.0114, 0.0113, 0.0081], + device='cuda:1'), out_proj_covar=tensor([1.0232e-04, 9.7613e-05, 1.1632e-04, 1.0515e-04, 1.2570e-04, 1.4840e-04, + 1.4720e-04, 1.0268e-04], device='cuda:1') +2023-03-21 00:09:26,590 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=46406.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:09:32,035 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 00:09:34,158 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.336e+02 2.112e+02 2.403e+02 2.943e+02 5.370e+02, threshold=4.806e+02, percent-clipped=0.0 +2023-03-21 00:09:39,344 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46430.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:09:41,786 INFO [train.py:901] (1/2) Epoch 17, batch 1250, loss[loss=0.1584, simple_loss=0.2359, pruned_loss=0.04043, over 7330.00 frames. ], tot_loss[loss=0.1593, simple_loss=0.2346, pruned_loss=0.04202, over 1438400.42 frames. ], batch size: 61, lr: 9.69e-03, grad_scale: 8.0 +2023-03-21 00:09:55,485 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 00:09:58,977 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 00:10:00,485 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 00:10:03,547 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=46478.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:10:07,021 INFO [train.py:901] (1/2) Epoch 17, batch 1300, loss[loss=0.1669, simple_loss=0.244, pruned_loss=0.04486, over 7279.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2351, pruned_loss=0.04232, over 1438249.32 frames. ], batch size: 68, lr: 9.69e-03, grad_scale: 8.0 +2023-03-21 00:10:23,647 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46515.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:10:25,046 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 00:10:26,035 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.425e+02 2.102e+02 2.512e+02 2.983e+02 5.030e+02, threshold=5.025e+02, percent-clipped=2.0 +2023-03-21 00:10:27,086 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 00:10:30,634 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 00:10:33,720 INFO [train.py:901] (1/2) Epoch 17, batch 1350, loss[loss=0.1625, simple_loss=0.2339, pruned_loss=0.04555, over 7360.00 frames. ], tot_loss[loss=0.1598, simple_loss=0.235, pruned_loss=0.04232, over 1439144.23 frames. ], batch size: 51, lr: 9.68e-03, grad_scale: 8.0 +2023-03-21 00:10:40,770 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 00:10:41,895 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2574, 3.7687, 3.8170, 3.8858, 3.7851, 3.8748, 4.1565, 3.6654], + device='cuda:1'), covar=tensor([0.0138, 0.0149, 0.0146, 0.0139, 0.0376, 0.0114, 0.0131, 0.0152], + device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0076, 0.0076, 0.0066, 0.0133, 0.0087, 0.0081, 0.0082], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:10:46,966 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8936, 3.3716, 2.7125, 3.0164, 2.9949, 2.3276, 3.1945, 2.9820], + device='cuda:1'), covar=tensor([0.1465, 0.0708, 0.1047, 0.1432, 0.1625, 0.1195, 0.1385, 0.1265], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0043, 0.0049, 0.0043, 0.0043, 0.0043, 0.0044, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:10:47,916 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=46563.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:10:50,475 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1368, 4.7004, 4.6907, 4.5987, 4.6055, 4.2990, 4.7269, 4.6364], + device='cuda:1'), covar=tensor([0.0471, 0.0398, 0.0436, 0.0530, 0.0303, 0.0349, 0.0369, 0.0384], + device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0208, 0.0150, 0.0153, 0.0127, 0.0189, 0.0162, 0.0122], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:10:59,826 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0419, 2.9982, 2.7199, 2.7999, 2.3137, 2.3095, 2.9890, 2.2407], + device='cuda:1'), covar=tensor([0.0311, 0.0364, 0.0372, 0.0378, 0.0436, 0.0592, 0.0435, 0.1115], + device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0316, 0.0255, 0.0339, 0.0305, 0.0306, 0.0321, 0.0295], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 00:11:00,133 INFO [train.py:901] (1/2) Epoch 17, batch 1400, loss[loss=0.1705, simple_loss=0.2499, pruned_loss=0.04558, over 7145.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.235, pruned_loss=0.04222, over 1438906.42 frames. ], batch size: 98, lr: 9.68e-03, grad_scale: 8.0 +2023-03-21 00:11:05,297 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46595.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:11:09,134 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46602.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:11:10,724 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8970, 3.2890, 2.6797, 4.0989, 1.8265, 3.8462, 1.7773, 3.3461], + device='cuda:1'), covar=tensor([0.0071, 0.0456, 0.1337, 0.0079, 0.4323, 0.0110, 0.1329, 0.0218], + device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0258, 0.0289, 0.0165, 0.0272, 0.0174, 0.0257, 0.0213], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 00:11:14,087 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 00:11:17,906 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.103e+02 2.396e+02 2.900e+02 8.640e+02, threshold=4.792e+02, percent-clipped=2.0 +2023-03-21 00:11:23,937 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46632.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:11:25,398 INFO [train.py:901] (1/2) Epoch 17, batch 1450, loss[loss=0.1566, simple_loss=0.2366, pruned_loss=0.03829, over 7220.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2346, pruned_loss=0.04213, over 1439958.07 frames. ], batch size: 45, lr: 9.67e-03, grad_scale: 8.0 +2023-03-21 00:11:36,068 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46656.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 00:11:38,522 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 00:11:40,140 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46663.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:11:51,668 INFO [train.py:901] (1/2) Epoch 17, batch 1500, loss[loss=0.1452, simple_loss=0.2192, pruned_loss=0.03562, over 7144.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2347, pruned_loss=0.04235, over 1439317.40 frames. ], batch size: 41, lr: 9.67e-03, grad_scale: 8.0 +2023-03-21 00:11:54,199 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 00:11:57,048 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.01 vs. limit=2.0 +2023-03-21 00:11:58,340 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1216, 3.0471, 2.0244, 3.6344, 2.3723, 3.1398, 1.6194, 1.8989], + device='cuda:1'), covar=tensor([0.0301, 0.0558, 0.2054, 0.0410, 0.0441, 0.0389, 0.3003, 0.1646], + device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0232, 0.0297, 0.0242, 0.0256, 0.0247, 0.0260, 0.0281], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 00:12:09,335 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.385e+02 2.167e+02 2.620e+02 3.081e+02 7.014e+02, threshold=5.239e+02, percent-clipped=3.0 +2023-03-21 00:12:16,862 INFO [train.py:901] (1/2) Epoch 17, batch 1550, loss[loss=0.1617, simple_loss=0.2351, pruned_loss=0.04421, over 7346.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2352, pruned_loss=0.04229, over 1441208.86 frames. ], batch size: 63, lr: 9.66e-03, grad_scale: 8.0 +2023-03-21 00:12:18,294 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 00:12:19,663 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-21 00:12:21,406 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7364, 2.9201, 2.4496, 2.9658, 2.8356, 2.3659, 3.1040, 2.6700], + device='cuda:1'), covar=tensor([0.0548, 0.1073, 0.1123, 0.0694, 0.1058, 0.0753, 0.0563, 0.1010], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0045, 0.0050, 0.0045, 0.0043, 0.0044, 0.0045, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:12:23,794 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-21 00:12:43,159 INFO [train.py:901] (1/2) Epoch 17, batch 1600, loss[loss=0.1188, simple_loss=0.1914, pruned_loss=0.02308, over 7161.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2353, pruned_loss=0.04249, over 1442482.73 frames. ], batch size: 39, lr: 9.66e-03, grad_scale: 8.0 +2023-03-21 00:12:43,332 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9012, 2.3729, 1.9003, 2.6105, 2.2172, 2.2687, 1.8620, 2.3252], + device='cuda:1'), covar=tensor([0.1654, 0.0706, 0.2880, 0.0507, 0.0086, 0.0072, 0.0144, 0.0188], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0234, 0.0267, 0.0261, 0.0147, 0.0140, 0.0170, 0.0189], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:12:51,685 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 00:12:52,707 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 00:12:55,695 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 00:12:56,346 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46810.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:13:01,325 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.076e+02 2.534e+02 3.354e+02 1.323e+03, threshold=5.067e+02, percent-clipped=3.0 +2023-03-21 00:13:02,537 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3565, 3.6624, 4.1612, 4.1404, 4.1962, 4.1570, 4.0963, 4.0298], + device='cuda:1'), covar=tensor([0.0022, 0.0085, 0.0032, 0.0032, 0.0027, 0.0030, 0.0027, 0.0052], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0052, 0.0046, 0.0044, 0.0044, 0.0047, 0.0045, 0.0057], + device='cuda:1'), out_proj_covar=tensor([8.2257e-05, 1.3169e-04, 1.1079e-04, 9.9141e-05, 9.8976e-05, 1.0255e-04, + 1.1062e-04, 1.2925e-04], device='cuda:1') +2023-03-21 00:13:04,115 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4285, 1.0501, 1.4485, 1.8351, 1.5505, 2.0050, 1.7054, 1.8401], + device='cuda:1'), covar=tensor([0.1412, 0.2931, 0.0864, 0.0775, 0.1591, 0.2322, 0.1341, 0.1708], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0056, 0.0038, 0.0039, 0.0042, 0.0043, 0.0059, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 00:13:05,003 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 00:13:08,993 INFO [train.py:901] (1/2) Epoch 17, batch 1650, loss[loss=0.1808, simple_loss=0.259, pruned_loss=0.05135, over 7160.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2345, pruned_loss=0.04196, over 1440608.64 frames. ], batch size: 98, lr: 9.65e-03, grad_scale: 8.0 +2023-03-21 00:13:09,005 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 00:13:17,796 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 00:13:26,055 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-03-21 00:13:28,461 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46871.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:13:34,817 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 00:13:35,151 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 00:13:35,337 INFO [train.py:901] (1/2) Epoch 17, batch 1700, loss[loss=0.1707, simple_loss=0.243, pruned_loss=0.04919, over 7358.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2351, pruned_loss=0.04233, over 1439615.37 frames. ], batch size: 63, lr: 9.65e-03, grad_scale: 8.0 +2023-03-21 00:13:39,481 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 00:13:50,343 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 00:13:52,746 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.281e+02 2.347e+02 2.673e+02 3.126e+02 7.111e+02, threshold=5.346e+02, percent-clipped=5.0 +2023-03-21 00:14:00,279 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46932.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:14:01,714 INFO [train.py:901] (1/2) Epoch 17, batch 1750, loss[loss=0.1866, simple_loss=0.2613, pruned_loss=0.05596, over 7278.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2352, pruned_loss=0.04237, over 1438161.58 frames. ], batch size: 70, lr: 9.64e-03, grad_scale: 8.0 +2023-03-21 00:14:09,816 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46951.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 00:14:13,358 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46958.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:14:15,790 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 00:14:16,794 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 00:14:24,196 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=46980.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:14:25,283 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1115, 2.4933, 1.9378, 2.7018, 2.5496, 2.2238, 2.0399, 2.5997], + device='cuda:1'), covar=tensor([0.1874, 0.0781, 0.2967, 0.0528, 0.0128, 0.0079, 0.0146, 0.0220], + device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0239, 0.0272, 0.0265, 0.0149, 0.0142, 0.0173, 0.0193], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:14:26,628 INFO [train.py:901] (1/2) Epoch 17, batch 1800, loss[loss=0.1756, simple_loss=0.2492, pruned_loss=0.05103, over 6701.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2352, pruned_loss=0.04241, over 1440504.79 frames. ], batch size: 106, lr: 9.64e-03, grad_scale: 8.0 +2023-03-21 00:14:37,548 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 00:14:45,779 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.279e+02 2.122e+02 2.479e+02 3.047e+02 5.778e+02, threshold=4.959e+02, percent-clipped=1.0 +2023-03-21 00:14:51,967 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 00:14:53,490 INFO [train.py:901] (1/2) Epoch 17, batch 1850, loss[loss=0.1536, simple_loss=0.2322, pruned_loss=0.03747, over 7272.00 frames. ], tot_loss[loss=0.1589, simple_loss=0.2343, pruned_loss=0.04177, over 1439587.33 frames. ], batch size: 52, lr: 9.63e-03, grad_scale: 8.0 +2023-03-21 00:15:01,534 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 00:15:13,717 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9853, 3.0598, 2.6307, 2.9415, 2.5510, 2.4542, 3.0961, 2.9531], + device='cuda:1'), covar=tensor([0.0958, 0.1473, 0.0848, 0.1272, 0.3440, 0.0857, 0.1657, 0.0970], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0046, 0.0051, 0.0045, 0.0045, 0.0044, 0.0046, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:15:17,645 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 00:15:18,604 INFO [train.py:901] (1/2) Epoch 17, batch 1900, loss[loss=0.1456, simple_loss=0.2166, pruned_loss=0.03728, over 7179.00 frames. ], tot_loss[loss=0.159, simple_loss=0.2345, pruned_loss=0.04175, over 1439851.87 frames. ], batch size: 39, lr: 9.63e-03, grad_scale: 8.0 +2023-03-21 00:15:22,272 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1432, 1.5699, 1.2463, 1.2100, 1.4627, 1.1337, 1.3286, 1.1240], + device='cuda:1'), covar=tensor([0.0131, 0.0092, 0.0248, 0.0159, 0.0156, 0.0168, 0.0111, 0.0129], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0021, 0.0021, 0.0021, 0.0022, 0.0021, 0.0023, 0.0028], + device='cuda:1'), out_proj_covar=tensor([2.8050e-05, 2.4347e-05, 2.4753e-05, 2.4163e-05, 2.6127e-05, 2.3951e-05, + 2.5952e-05, 3.3128e-05], device='cuda:1') +2023-03-21 00:15:37,427 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.393e+02 2.079e+02 2.464e+02 3.016e+02 5.174e+02, threshold=4.929e+02, percent-clipped=1.0 +2023-03-21 00:15:44,639 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 00:15:45,135 INFO [train.py:901] (1/2) Epoch 17, batch 1950, loss[loss=0.1541, simple_loss=0.2293, pruned_loss=0.03941, over 7346.00 frames. ], tot_loss[loss=0.1578, simple_loss=0.2332, pruned_loss=0.04121, over 1439059.62 frames. ], batch size: 51, lr: 9.62e-03, grad_scale: 16.0 +2023-03-21 00:15:45,428 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.02 vs. limit=5.0 +2023-03-21 00:15:54,704 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 00:15:59,226 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 00:15:59,782 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 00:16:00,860 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47166.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:16:02,387 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4991, 3.6870, 3.4605, 3.5774, 3.4621, 3.6372, 3.9867, 3.9563], + device='cuda:1'), covar=tensor([0.0272, 0.0179, 0.0260, 0.0237, 0.0331, 0.0360, 0.0238, 0.0220], + device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0110, 0.0104, 0.0111, 0.0104, 0.0093, 0.0090, 0.0089], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:16:10,988 INFO [train.py:901] (1/2) Epoch 17, batch 2000, loss[loss=0.1372, simple_loss=0.2156, pruned_loss=0.02943, over 7334.00 frames. ], tot_loss[loss=0.1581, simple_loss=0.2336, pruned_loss=0.04126, over 1440750.86 frames. ], batch size: 44, lr: 9.62e-03, grad_scale: 16.0 +2023-03-21 00:16:16,647 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 00:16:24,493 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3286, 2.6874, 2.7496, 2.3599, 2.5017, 2.3476, 2.5591, 1.8683], + device='cuda:1'), covar=tensor([0.0225, 0.0169, 0.0060, 0.0097, 0.0413, 0.0225, 0.0099, 0.0318], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0026, 0.0022, 0.0022, 0.0023, 0.0022, 0.0026, 0.0026], + device='cuda:1'), out_proj_covar=tensor([6.3056e-05, 6.4492e-05, 5.5790e-05, 5.4574e-05, 5.9944e-05, 5.6881e-05, + 6.3108e-05, 6.6616e-05], device='cuda:1') +2023-03-21 00:16:28,340 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 00:16:29,339 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.291e+02 2.033e+02 2.454e+02 2.931e+02 5.819e+02, threshold=4.907e+02, percent-clipped=3.0 +2023-03-21 00:16:36,530 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 00:16:37,042 INFO [train.py:901] (1/2) Epoch 17, batch 2050, loss[loss=0.1564, simple_loss=0.2283, pruned_loss=0.04222, over 7262.00 frames. ], tot_loss[loss=0.1577, simple_loss=0.233, pruned_loss=0.04119, over 1440838.96 frames. ], batch size: 45, lr: 9.61e-03, grad_scale: 16.0 +2023-03-21 00:16:40,256 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47241.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:16:45,220 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47251.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:16:48,635 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47258.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:17:03,232 INFO [train.py:901] (1/2) Epoch 17, batch 2100, loss[loss=0.1795, simple_loss=0.2468, pruned_loss=0.05614, over 7251.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2328, pruned_loss=0.0412, over 1441185.74 frames. ], batch size: 47, lr: 9.61e-03, grad_scale: 8.0 +2023-03-21 00:17:10,521 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=47299.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:17:11,448 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 00:17:12,091 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47302.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:17:13,996 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=47306.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:17:14,451 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 00:17:21,500 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.164e+02 2.514e+02 3.054e+02 5.580e+02, threshold=5.028e+02, percent-clipped=1.0 +2023-03-21 00:17:28,635 INFO [train.py:901] (1/2) Epoch 17, batch 2150, loss[loss=0.1534, simple_loss=0.2363, pruned_loss=0.03524, over 7309.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2323, pruned_loss=0.04111, over 1439043.87 frames. ], batch size: 83, lr: 9.60e-03, grad_scale: 8.0 +2023-03-21 00:17:30,280 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0865, 2.7469, 3.0806, 2.7753, 3.2012, 2.7600, 2.1688, 3.0130], + device='cuda:1'), covar=tensor([0.1492, 0.0668, 0.1487, 0.2488, 0.0946, 0.1465, 0.3563, 0.1614], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0049, 0.0039, 0.0040, 0.0038, 0.0036, 0.0055, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 00:17:52,392 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47379.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:17:55,210 INFO [train.py:901] (1/2) Epoch 17, batch 2200, loss[loss=0.1455, simple_loss=0.2188, pruned_loss=0.03611, over 7274.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2328, pruned_loss=0.04095, over 1441402.78 frames. ], batch size: 77, lr: 9.60e-03, grad_scale: 8.0 +2023-03-21 00:18:00,294 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 00:18:13,751 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.428e+02 2.138e+02 2.450e+02 3.196e+02 6.599e+02, threshold=4.899e+02, percent-clipped=2.0 +2023-03-21 00:18:17,930 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7276, 2.1312, 2.3091, 2.0054, 2.0725, 1.7476, 1.9112, 1.5277], + device='cuda:1'), covar=tensor([0.0432, 0.0547, 0.0172, 0.0184, 0.0302, 0.0319, 0.0202, 0.0315], + device='cuda:1'), in_proj_covar=tensor([0.0026, 0.0026, 0.0023, 0.0023, 0.0023, 0.0022, 0.0026, 0.0027], + device='cuda:1'), out_proj_covar=tensor([6.4669e-05, 6.5053e-05, 5.6268e-05, 5.5740e-05, 6.0437e-05, 5.7229e-05, + 6.3582e-05, 6.7492e-05], device='cuda:1') +2023-03-21 00:18:20,804 INFO [train.py:901] (1/2) Epoch 17, batch 2250, loss[loss=0.1677, simple_loss=0.2451, pruned_loss=0.04515, over 7258.00 frames. ], tot_loss[loss=0.1567, simple_loss=0.232, pruned_loss=0.0407, over 1440247.45 frames. ], batch size: 55, lr: 9.59e-03, grad_scale: 8.0 +2023-03-21 00:18:23,487 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47440.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:18:35,191 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 00:18:35,734 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 00:18:37,810 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47466.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:18:39,040 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-21 00:18:39,601 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-21 00:18:47,230 INFO [train.py:901] (1/2) Epoch 17, batch 2300, loss[loss=0.1626, simple_loss=0.2366, pruned_loss=0.04428, over 7359.00 frames. ], tot_loss[loss=0.1568, simple_loss=0.2323, pruned_loss=0.04066, over 1441780.66 frames. ], batch size: 73, lr: 9.59e-03, grad_scale: 8.0 +2023-03-21 00:18:48,220 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 00:19:01,777 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=47514.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:19:05,248 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.279e+02 2.078e+02 2.405e+02 2.974e+02 9.251e+02, threshold=4.809e+02, percent-clipped=4.0 +2023-03-21 00:19:07,474 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5614, 3.1846, 2.3434, 3.9732, 2.9917, 3.4285, 1.7966, 2.2224], + device='cuda:1'), covar=tensor([0.0232, 0.0559, 0.1582, 0.0373, 0.0275, 0.0419, 0.2273, 0.1371], + device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0239, 0.0299, 0.0245, 0.0255, 0.0254, 0.0263, 0.0283], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 00:19:12,299 INFO [train.py:901] (1/2) Epoch 17, batch 2350, loss[loss=0.1704, simple_loss=0.2398, pruned_loss=0.05052, over 7282.00 frames. ], tot_loss[loss=0.1571, simple_loss=0.2325, pruned_loss=0.0409, over 1440272.38 frames. ], batch size: 66, lr: 9.58e-03, grad_scale: 8.0 +2023-03-21 00:19:34,534 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 00:19:38,527 INFO [train.py:901] (1/2) Epoch 17, batch 2400, loss[loss=0.149, simple_loss=0.2302, pruned_loss=0.03394, over 7307.00 frames. ], tot_loss[loss=0.1568, simple_loss=0.2323, pruned_loss=0.04066, over 1441361.64 frames. ], batch size: 86, lr: 9.58e-03, grad_scale: 8.0 +2023-03-21 00:19:41,102 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 00:19:44,637 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47597.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:19:47,475 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2544, 4.4101, 4.1547, 4.4565, 4.1083, 4.4844, 4.6684, 4.6904], + device='cuda:1'), covar=tensor([0.0193, 0.0143, 0.0182, 0.0124, 0.0237, 0.0152, 0.0189, 0.0155], + device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0111, 0.0103, 0.0109, 0.0103, 0.0092, 0.0091, 0.0087], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:19:47,985 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7898, 5.2757, 5.2592, 5.2838, 5.0613, 4.7211, 5.3439, 5.0702], + device='cuda:1'), covar=tensor([0.0454, 0.0427, 0.0481, 0.0394, 0.0365, 0.0347, 0.0370, 0.0601], + device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0216, 0.0155, 0.0156, 0.0130, 0.0197, 0.0166, 0.0127], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:19:49,222 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 +2023-03-21 00:19:52,378 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 00:19:55,394 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 00:19:56,803 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.471e+02 2.109e+02 2.488e+02 3.025e+02 6.588e+02, threshold=4.976e+02, percent-clipped=2.0 +2023-03-21 00:20:05,128 INFO [train.py:901] (1/2) Epoch 17, batch 2450, loss[loss=0.1599, simple_loss=0.2378, pruned_loss=0.04103, over 7329.00 frames. ], tot_loss[loss=0.1565, simple_loss=0.2321, pruned_loss=0.04041, over 1441484.81 frames. ], batch size: 61, lr: 9.57e-03, grad_scale: 8.0 +2023-03-21 00:20:22,037 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 00:20:27,364 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 +2023-03-21 00:20:30,618 INFO [train.py:901] (1/2) Epoch 17, batch 2500, loss[loss=0.1698, simple_loss=0.25, pruned_loss=0.04476, over 7352.00 frames. ], tot_loss[loss=0.1572, simple_loss=0.2329, pruned_loss=0.04074, over 1441638.32 frames. ], batch size: 63, lr: 9.57e-03, grad_scale: 8.0 +2023-03-21 00:20:35,850 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1322, 2.8252, 2.9349, 2.9550, 3.1525, 2.9098, 2.5368, 3.3352], + device='cuda:1'), covar=tensor([0.1724, 0.0549, 0.2448, 0.2967, 0.1171, 0.1397, 0.2442, 0.0995], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0050, 0.0040, 0.0040, 0.0039, 0.0036, 0.0055, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 00:20:47,581 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 00:20:50,038 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.424e+02 2.240e+02 2.665e+02 3.286e+02 7.455e+02, threshold=5.331e+02, percent-clipped=3.0 +2023-03-21 00:20:57,338 INFO [train.py:901] (1/2) Epoch 17, batch 2550, loss[loss=0.1778, simple_loss=0.2508, pruned_loss=0.05242, over 7227.00 frames. ], tot_loss[loss=0.1575, simple_loss=0.2333, pruned_loss=0.0409, over 1442714.94 frames. ], batch size: 93, lr: 9.56e-03, grad_scale: 8.0 +2023-03-21 00:20:57,422 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47735.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:20:59,035 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47738.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:21:02,141 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3690, 2.8219, 2.1923, 3.7232, 2.5727, 3.0250, 1.6314, 2.0608], + device='cuda:1'), covar=tensor([0.0285, 0.0589, 0.2080, 0.0483, 0.0349, 0.0428, 0.2922, 0.1995], + device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0243, 0.0304, 0.0247, 0.0259, 0.0257, 0.0267, 0.0286], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 00:21:09,283 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9738, 2.5579, 1.9677, 3.0944, 2.6910, 2.6402, 2.3008, 2.4322], + device='cuda:1'), covar=tensor([0.1759, 0.0661, 0.2733, 0.0434, 0.0142, 0.0086, 0.0176, 0.0196], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0231, 0.0264, 0.0257, 0.0146, 0.0141, 0.0173, 0.0191], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:21:22,595 INFO [train.py:901] (1/2) Epoch 17, batch 2600, loss[loss=0.166, simple_loss=0.2478, pruned_loss=0.04212, over 7271.00 frames. ], tot_loss[loss=0.1577, simple_loss=0.2335, pruned_loss=0.04093, over 1442117.93 frames. ], batch size: 64, lr: 9.56e-03, grad_scale: 8.0 +2023-03-21 00:21:24,222 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47788.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:21:29,826 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47799.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:21:34,964 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47809.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 00:21:41,944 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.493e+02 2.119e+02 2.477e+02 3.299e+02 7.952e+02, threshold=4.955e+02, percent-clipped=2.0 +2023-03-21 00:21:44,513 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1438, 1.4122, 1.3998, 1.2920, 1.4377, 1.3518, 1.2546, 0.9895], + device='cuda:1'), covar=tensor([0.0109, 0.0102, 0.0128, 0.0133, 0.0090, 0.0071, 0.0228, 0.0115], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0022, 0.0022, 0.0021, 0.0022, 0.0021, 0.0023, 0.0029], + device='cuda:1'), out_proj_covar=tensor([2.9038e-05, 2.4867e-05, 2.5481e-05, 2.4171e-05, 2.6301e-05, 2.4175e-05, + 2.6950e-05, 3.3529e-05], device='cuda:1') +2023-03-21 00:21:48,754 INFO [train.py:901] (1/2) Epoch 17, batch 2650, loss[loss=0.1883, simple_loss=0.2573, pruned_loss=0.05964, over 7268.00 frames. ], tot_loss[loss=0.158, simple_loss=0.2338, pruned_loss=0.04107, over 1444182.51 frames. ], batch size: 70, lr: 9.55e-03, grad_scale: 8.0 +2023-03-21 00:21:55,775 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47849.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:22:06,161 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47870.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 00:22:13,285 INFO [train.py:901] (1/2) Epoch 17, batch 2700, loss[loss=0.1687, simple_loss=0.2324, pruned_loss=0.05254, over 7251.00 frames. ], tot_loss[loss=0.1584, simple_loss=0.2341, pruned_loss=0.0414, over 1443113.94 frames. ], batch size: 47, lr: 9.55e-03, grad_scale: 8.0 +2023-03-21 00:22:19,233 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47897.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:22:30,939 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 2.095e+02 2.437e+02 3.021e+02 7.441e+02, threshold=4.873e+02, percent-clipped=1.0 +2023-03-21 00:22:37,934 INFO [train.py:901] (1/2) Epoch 17, batch 2750, loss[loss=0.1408, simple_loss=0.2202, pruned_loss=0.0307, over 7364.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.2343, pruned_loss=0.04139, over 1445298.43 frames. ], batch size: 51, lr: 9.54e-03, grad_scale: 8.0 +2023-03-21 00:22:43,008 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=47945.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:23:02,679 INFO [train.py:901] (1/2) Epoch 17, batch 2800, loss[loss=0.1975, simple_loss=0.2713, pruned_loss=0.06191, over 6784.00 frames. ], tot_loss[loss=0.1582, simple_loss=0.2338, pruned_loss=0.04124, over 1444025.90 frames. ], batch size: 107, lr: 9.54e-03, grad_scale: 8.0 +2023-03-21 00:23:06,782 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2887, 4.4045, 4.1827, 4.3735, 4.1621, 4.1976, 4.5719, 4.6540], + device='cuda:1'), covar=tensor([0.0263, 0.0277, 0.0295, 0.0276, 0.0434, 0.0464, 0.0291, 0.0245], + device='cuda:1'), in_proj_covar=tensor([0.0108, 0.0110, 0.0102, 0.0108, 0.0103, 0.0091, 0.0089, 0.0084], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:23:31,983 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 00:23:32,036 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 00:23:32,064 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 00:23:32,082 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 00:23:32,251 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 00:23:32,267 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 00:23:32,610 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 00:23:32,705 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 00:23:32,783 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 00:23:32,797 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 00:23:32,815 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 00:23:32,993 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 00:23:32,998 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 00:23:33,010 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 00:23:33,146 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 00:23:33,149 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 00:23:33,206 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 00:23:40,433 INFO [train.py:901] (1/2) Epoch 18, batch 0, loss[loss=0.1534, simple_loss=0.2272, pruned_loss=0.0398, over 7264.00 frames. ], tot_loss[loss=0.1534, simple_loss=0.2272, pruned_loss=0.0398, over 7264.00 frames. ], batch size: 64, lr: 9.28e-03, grad_scale: 8.0 +2023-03-21 00:23:40,433 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 00:24:06,020 INFO [train.py:935] (1/2) Epoch 18, validation: loss=0.1665, simple_loss=0.253, pruned_loss=0.03999, over 1622729.00 frames. +2023-03-21 00:24:06,020 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 00:24:06,719 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1890, 3.0787, 2.9822, 3.0534, 2.5276, 2.6262, 3.2557, 2.3874], + device='cuda:1'), covar=tensor([0.0399, 0.0242, 0.0329, 0.0399, 0.0410, 0.0530, 0.0382, 0.1061], + device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0313, 0.0254, 0.0339, 0.0304, 0.0302, 0.0321, 0.0289], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 00:24:12,046 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.469e+02 1.972e+02 2.348e+02 2.792e+02 4.861e+02, threshold=4.696e+02, percent-clipped=0.0 +2023-03-21 00:24:13,102 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 00:24:19,080 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48035.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:24:21,178 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3250, 3.2608, 2.2410, 3.8521, 2.9942, 3.3441, 1.7699, 2.0989], + device='cuda:1'), covar=tensor([0.0296, 0.0855, 0.2014, 0.0388, 0.0372, 0.0734, 0.2666, 0.1756], + device='cuda:1'), in_proj_covar=tensor([0.0258, 0.0240, 0.0301, 0.0245, 0.0255, 0.0256, 0.0265, 0.0282], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 00:24:24,792 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 00:24:27,449 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8065, 3.2404, 2.5664, 4.0373, 1.4945, 3.7709, 1.6307, 2.9750], + device='cuda:1'), covar=tensor([0.0070, 0.0678, 0.1535, 0.0082, 0.4651, 0.0129, 0.1148, 0.0193], + device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0266, 0.0294, 0.0172, 0.0277, 0.0184, 0.0264, 0.0220], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 00:24:31,968 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 00:24:32,462 INFO [train.py:901] (1/2) Epoch 18, batch 50, loss[loss=0.1711, simple_loss=0.249, pruned_loss=0.04663, over 7342.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2358, pruned_loss=0.04294, over 327402.03 frames. ], batch size: 54, lr: 9.28e-03, grad_scale: 8.0 +2023-03-21 00:24:34,540 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 00:24:36,677 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6319, 5.1016, 5.1816, 5.0901, 4.9425, 4.6988, 5.1868, 4.9754], + device='cuda:1'), covar=tensor([0.0405, 0.0325, 0.0295, 0.0388, 0.0259, 0.0273, 0.0259, 0.0390], + device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0214, 0.0155, 0.0157, 0.0128, 0.0196, 0.0164, 0.0126], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:24:37,137 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 00:24:38,260 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.2366, 4.7721, 4.7734, 5.2261, 5.1880, 5.1925, 4.5902, 4.6758], + device='cuda:1'), covar=tensor([0.0647, 0.2506, 0.2068, 0.0957, 0.0758, 0.1169, 0.0651, 0.0958], + device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0319, 0.0250, 0.0249, 0.0183, 0.0314, 0.0182, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:24:39,313 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48072.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:24:43,051 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.08 vs. limit=2.0 +2023-03-21 00:24:44,728 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=48083.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:24:50,389 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48094.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:24:53,865 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 00:24:54,413 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 00:24:57,973 INFO [train.py:901] (1/2) Epoch 18, batch 100, loss[loss=0.1256, simple_loss=0.1818, pruned_loss=0.03467, over 6015.00 frames. ], tot_loss[loss=0.1563, simple_loss=0.2313, pruned_loss=0.04067, over 567194.84 frames. ], batch size: 26, lr: 9.27e-03, grad_scale: 8.0 +2023-03-21 00:25:03,911 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.597e+02 2.080e+02 2.464e+02 3.092e+02 6.621e+02, threshold=4.927e+02, percent-clipped=4.0 +2023-03-21 00:25:10,735 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48133.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:25:15,377 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6079, 3.6689, 3.5934, 3.7451, 3.3974, 3.6524, 3.9820, 3.9105], + device='cuda:1'), covar=tensor([0.0231, 0.0176, 0.0220, 0.0164, 0.0358, 0.0384, 0.0197, 0.0204], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0109, 0.0100, 0.0107, 0.0101, 0.0090, 0.0088, 0.0084], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:25:16,880 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48144.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:25:24,276 INFO [train.py:901] (1/2) Epoch 18, batch 150, loss[loss=0.1278, simple_loss=0.1967, pruned_loss=0.02943, over 6950.00 frames. ], tot_loss[loss=0.1563, simple_loss=0.232, pruned_loss=0.04033, over 762150.81 frames. ], batch size: 35, lr: 9.27e-03, grad_scale: 8.0 +2023-03-21 00:25:27,371 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48165.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 00:25:49,726 INFO [train.py:901] (1/2) Epoch 18, batch 200, loss[loss=0.157, simple_loss=0.2366, pruned_loss=0.03869, over 7285.00 frames. ], tot_loss[loss=0.1569, simple_loss=0.2323, pruned_loss=0.04075, over 911875.85 frames. ], batch size: 77, lr: 9.26e-03, grad_scale: 8.0 +2023-03-21 00:25:51,913 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48212.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 00:25:55,327 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 00:25:56,913 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.110e+02 2.629e+02 3.068e+02 6.550e+02, threshold=5.258e+02, percent-clipped=1.0 +2023-03-21 00:25:59,975 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 00:26:06,480 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 00:26:10,267 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-03-21 00:26:15,922 INFO [train.py:901] (1/2) Epoch 18, batch 250, loss[loss=0.1534, simple_loss=0.2313, pruned_loss=0.03774, over 7312.00 frames. ], tot_loss[loss=0.1567, simple_loss=0.2324, pruned_loss=0.0405, over 1030602.66 frames. ], batch size: 83, lr: 9.26e-03, grad_scale: 8.0 +2023-03-21 00:26:19,479 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 00:26:23,139 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48273.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 00:26:40,404 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 00:26:42,365 INFO [train.py:901] (1/2) Epoch 18, batch 300, loss[loss=0.1422, simple_loss=0.2199, pruned_loss=0.03225, over 7326.00 frames. ], tot_loss[loss=0.157, simple_loss=0.2326, pruned_loss=0.04072, over 1121766.26 frames. ], batch size: 75, lr: 9.26e-03, grad_scale: 8.0 +2023-03-21 00:26:48,314 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.449e+02 2.135e+02 2.501e+02 2.825e+02 5.059e+02, threshold=5.003e+02, percent-clipped=0.0 +2023-03-21 00:26:48,851 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 00:26:49,962 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48324.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:26:52,452 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6389, 2.9306, 3.4032, 3.5225, 3.5130, 3.7074, 3.4737, 3.4158], + device='cuda:1'), covar=tensor([0.0030, 0.0102, 0.0046, 0.0037, 0.0039, 0.0026, 0.0053, 0.0060], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0050, 0.0044, 0.0041, 0.0042, 0.0045, 0.0043, 0.0055], + device='cuda:1'), out_proj_covar=tensor([7.8679e-05, 1.2517e-04, 1.0278e-04, 8.8619e-05, 9.3541e-05, 9.8273e-05, + 1.0396e-04, 1.2299e-04], device='cuda:1') +2023-03-21 00:27:07,224 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-03-21 00:27:07,447 INFO [train.py:901] (1/2) Epoch 18, batch 350, loss[loss=0.1603, simple_loss=0.2367, pruned_loss=0.04195, over 7301.00 frames. ], tot_loss[loss=0.1568, simple_loss=0.2324, pruned_loss=0.04057, over 1193341.95 frames. ], batch size: 49, lr: 9.25e-03, grad_scale: 8.0 +2023-03-21 00:27:09,818 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 +2023-03-21 00:27:15,788 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6756, 3.4073, 3.3899, 3.1047, 2.6673, 2.9088, 3.7318, 2.7206], + device='cuda:1'), covar=tensor([0.0236, 0.0290, 0.0318, 0.0308, 0.0447, 0.0602, 0.0322, 0.1071], + device='cuda:1'), in_proj_covar=tensor([0.0315, 0.0318, 0.0257, 0.0339, 0.0308, 0.0308, 0.0326, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 00:27:20,888 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48385.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:27:23,772 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 00:27:26,479 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48394.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:27:27,921 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48397.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:27:34,234 INFO [train.py:901] (1/2) Epoch 18, batch 400, loss[loss=0.1481, simple_loss=0.2223, pruned_loss=0.03695, over 7379.00 frames. ], tot_loss[loss=0.1552, simple_loss=0.2309, pruned_loss=0.03975, over 1245795.28 frames. ], batch size: 51, lr: 9.25e-03, grad_scale: 8.0 +2023-03-21 00:27:37,126 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-03-21 00:27:40,286 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.256e+02 1.977e+02 2.369e+02 2.933e+02 5.499e+02, threshold=4.738e+02, percent-clipped=2.0 +2023-03-21 00:27:41,469 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48423.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:27:43,860 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48428.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:27:45,986 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2952, 2.7031, 1.9167, 3.3415, 2.3760, 2.6426, 1.6649, 1.9128], + device='cuda:1'), covar=tensor([0.0265, 0.0630, 0.2207, 0.0410, 0.0453, 0.0443, 0.2796, 0.1717], + device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0240, 0.0298, 0.0244, 0.0259, 0.0255, 0.0263, 0.0282], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 00:27:50,871 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=48442.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:27:51,946 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48444.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:27:58,515 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0186, 2.4109, 1.7886, 3.0474, 2.5609, 2.7687, 2.4199, 2.3118], + device='cuda:1'), covar=tensor([0.1800, 0.0737, 0.3111, 0.0586, 0.0103, 0.0071, 0.0195, 0.0204], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0233, 0.0268, 0.0263, 0.0151, 0.0145, 0.0180, 0.0194], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:27:59,000 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48458.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:27:59,327 INFO [train.py:901] (1/2) Epoch 18, batch 450, loss[loss=0.1614, simple_loss=0.2396, pruned_loss=0.04158, over 7268.00 frames. ], tot_loss[loss=0.1564, simple_loss=0.2323, pruned_loss=0.04019, over 1291026.72 frames. ], batch size: 64, lr: 9.24e-03, grad_scale: 8.0 +2023-03-21 00:28:02,533 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48465.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 00:28:03,432 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 00:28:04,408 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 00:28:13,410 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48484.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:28:17,238 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=48492.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:28:25,774 INFO [train.py:901] (1/2) Epoch 18, batch 500, loss[loss=0.1258, simple_loss=0.1999, pruned_loss=0.02588, over 7169.00 frames. ], tot_loss[loss=0.1564, simple_loss=0.2324, pruned_loss=0.04018, over 1325471.21 frames. ], batch size: 39, lr: 9.24e-03, grad_scale: 8.0 +2023-03-21 00:28:27,844 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=48513.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 00:28:31,852 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.337e+02 2.257e+02 2.553e+02 3.004e+02 5.113e+02, threshold=5.107e+02, percent-clipped=1.0 +2023-03-21 00:28:33,029 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2332, 0.7997, 1.5388, 1.6194, 1.3326, 1.6796, 1.2915, 1.4765], + device='cuda:1'), covar=tensor([0.1343, 0.5349, 0.1307, 0.0911, 0.1596, 0.2044, 0.1473, 0.2965], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0059, 0.0041, 0.0040, 0.0041, 0.0045, 0.0063, 0.0043], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 00:28:37,106 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.9647, 1.2362, 1.1712, 1.2872, 1.2923, 1.1761, 1.2620, 0.8436], + device='cuda:1'), covar=tensor([0.0120, 0.0079, 0.0112, 0.0069, 0.0107, 0.0063, 0.0109, 0.0105], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0023, 0.0022, 0.0022, 0.0023, 0.0022, 0.0023, 0.0029], + device='cuda:1'), out_proj_covar=tensor([2.8952e-05, 2.5732e-05, 2.5346e-05, 2.4768e-05, 2.7183e-05, 2.4468e-05, + 2.6891e-05, 3.4340e-05], device='cuda:1') +2023-03-21 00:28:37,961 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 00:28:39,512 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 00:28:40,008 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 00:28:42,001 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 00:28:46,464 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 00:28:50,913 INFO [train.py:901] (1/2) Epoch 18, batch 550, loss[loss=0.1591, simple_loss=0.2326, pruned_loss=0.04279, over 7258.00 frames. ], tot_loss[loss=0.1562, simple_loss=0.2321, pruned_loss=0.04015, over 1350886.15 frames. ], batch size: 55, lr: 9.23e-03, grad_scale: 8.0 +2023-03-21 00:28:56,916 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48568.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 00:28:59,799 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 00:29:07,913 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 00:29:11,394 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 00:29:17,778 INFO [train.py:901] (1/2) Epoch 18, batch 600, loss[loss=0.1641, simple_loss=0.2453, pruned_loss=0.04143, over 7346.00 frames. ], tot_loss[loss=0.1574, simple_loss=0.2332, pruned_loss=0.04086, over 1373158.73 frames. ], batch size: 54, lr: 9.23e-03, grad_scale: 8.0 +2023-03-21 00:29:18,316 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 00:29:22,459 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6951, 2.7566, 1.8269, 3.3972, 2.4384, 2.8704, 1.5052, 1.8325], + device='cuda:1'), covar=tensor([0.0320, 0.0707, 0.2560, 0.0526, 0.0466, 0.0532, 0.3234, 0.1855], + device='cuda:1'), in_proj_covar=tensor([0.0260, 0.0244, 0.0305, 0.0248, 0.0263, 0.0258, 0.0270, 0.0286], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 00:29:23,768 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.190e+02 2.516e+02 3.083e+02 5.159e+02, threshold=5.033e+02, percent-clipped=1.0 +2023-03-21 00:29:34,363 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 00:29:42,726 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1499, 4.6804, 4.7753, 4.6216, 4.6332, 4.1879, 4.7707, 4.5799], + device='cuda:1'), covar=tensor([0.0477, 0.0410, 0.0341, 0.0433, 0.0302, 0.0360, 0.0281, 0.0526], + device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0214, 0.0152, 0.0154, 0.0126, 0.0195, 0.0162, 0.0125], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:29:43,690 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 00:29:44,205 INFO [train.py:901] (1/2) Epoch 18, batch 650, loss[loss=0.1347, simple_loss=0.2023, pruned_loss=0.03359, over 6941.00 frames. ], tot_loss[loss=0.1569, simple_loss=0.2329, pruned_loss=0.04045, over 1391047.70 frames. ], batch size: 35, lr: 9.22e-03, grad_scale: 8.0 +2023-03-21 00:29:53,311 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7035, 3.7606, 3.1651, 3.1514, 2.8682, 2.3287, 1.7139, 3.6860], + device='cuda:1'), covar=tensor([0.0032, 0.0028, 0.0084, 0.0069, 0.0109, 0.0358, 0.0451, 0.0047], + device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0071, 0.0088, 0.0078, 0.0098, 0.0114, 0.0113, 0.0083], + device='cuda:1'), out_proj_covar=tensor([1.0281e-04, 9.8065e-05, 1.1390e-04, 1.0431e-04, 1.2413e-04, 1.4626e-04, + 1.4637e-04, 1.0419e-04], device='cuda:1') +2023-03-21 00:29:54,786 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48680.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:29:59,666 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 00:30:07,680 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 00:30:09,160 INFO [train.py:901] (1/2) Epoch 18, batch 700, loss[loss=0.148, simple_loss=0.2284, pruned_loss=0.03383, over 7135.00 frames. ], tot_loss[loss=0.1577, simple_loss=0.2336, pruned_loss=0.04091, over 1401576.89 frames. ], batch size: 41, lr: 9.22e-03, grad_scale: 8.0 +2023-03-21 00:30:15,214 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.344e+02 2.059e+02 2.384e+02 3.008e+02 6.092e+02, threshold=4.769e+02, percent-clipped=1.0 +2023-03-21 00:30:18,855 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48728.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:30:19,900 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([0.9119, 1.1903, 1.2369, 1.2411, 1.4362, 1.2625, 1.1813, 0.8965], + device='cuda:1'), covar=tensor([0.0130, 0.0197, 0.0176, 0.0192, 0.0081, 0.0110, 0.0174, 0.0149], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0023, 0.0022, 0.0022, 0.0023, 0.0022, 0.0024, 0.0030], + device='cuda:1'), out_proj_covar=tensor([2.9240e-05, 2.6418e-05, 2.5507e-05, 2.5317e-05, 2.7439e-05, 2.4935e-05, + 2.8156e-05, 3.4756e-05], device='cuda:1') +2023-03-21 00:30:32,348 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48753.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:30:32,447 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3678, 3.1205, 3.1027, 3.1500, 2.5586, 2.7334, 3.2303, 2.5766], + device='cuda:1'), covar=tensor([0.0289, 0.0257, 0.0354, 0.0286, 0.0331, 0.0530, 0.0378, 0.1120], + device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0323, 0.0258, 0.0343, 0.0307, 0.0305, 0.0329, 0.0295], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 00:30:32,799 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 00:30:33,287 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 00:30:35,267 INFO [train.py:901] (1/2) Epoch 18, batch 750, loss[loss=0.1356, simple_loss=0.2141, pruned_loss=0.0286, over 7327.00 frames. ], tot_loss[loss=0.157, simple_loss=0.2331, pruned_loss=0.04051, over 1412294.07 frames. ], batch size: 44, lr: 9.21e-03, grad_scale: 8.0 +2023-03-21 00:30:43,989 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=48776.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:30:45,512 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48779.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:30:47,455 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 00:30:51,505 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 00:30:56,643 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 +2023-03-21 00:30:57,825 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 00:30:59,309 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 00:30:59,919 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9475, 4.0968, 3.9595, 4.0489, 3.7897, 4.0968, 4.4117, 4.4323], + device='cuda:1'), covar=tensor([0.0218, 0.0177, 0.0199, 0.0165, 0.0278, 0.0215, 0.0252, 0.0185], + device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0110, 0.0100, 0.0109, 0.0101, 0.0091, 0.0088, 0.0085], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:31:00,842 INFO [train.py:901] (1/2) Epoch 18, batch 800, loss[loss=0.1251, simple_loss=0.2114, pruned_loss=0.0194, over 7107.00 frames. ], tot_loss[loss=0.1574, simple_loss=0.2336, pruned_loss=0.04057, over 1420910.31 frames. ], batch size: 41, lr: 9.21e-03, grad_scale: 8.0 +2023-03-21 00:31:03,652 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 00:31:07,234 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.247e+02 2.109e+02 2.468e+02 2.979e+02 5.290e+02, threshold=4.936e+02, percent-clipped=1.0 +2023-03-21 00:31:07,358 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48821.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:31:10,840 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 00:31:16,461 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48838.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:31:26,876 INFO [train.py:901] (1/2) Epoch 18, batch 850, loss[loss=0.1359, simple_loss=0.2115, pruned_loss=0.03012, over 7139.00 frames. ], tot_loss[loss=0.1565, simple_loss=0.2327, pruned_loss=0.04015, over 1426974.96 frames. ], batch size: 39, lr: 9.20e-03, grad_scale: 8.0 +2023-03-21 00:31:29,925 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 00:31:30,393 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 00:31:31,435 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48868.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 00:31:34,734 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 00:31:37,829 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 00:31:38,452 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48882.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:31:46,830 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48899.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:31:52,977 INFO [train.py:901] (1/2) Epoch 18, batch 900, loss[loss=0.123, simple_loss=0.1959, pruned_loss=0.02506, over 6936.00 frames. ], tot_loss[loss=0.1578, simple_loss=0.2339, pruned_loss=0.04091, over 1432559.73 frames. ], batch size: 35, lr: 9.20e-03, grad_scale: 8.0 +2023-03-21 00:31:56,429 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=48916.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 00:31:58,845 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.424e+02 2.181e+02 2.751e+02 3.267e+02 5.780e+02, threshold=5.502e+02, percent-clipped=4.0 +2023-03-21 00:32:15,674 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 00:32:18,144 INFO [train.py:901] (1/2) Epoch 18, batch 950, loss[loss=0.1836, simple_loss=0.2503, pruned_loss=0.05849, over 7272.00 frames. ], tot_loss[loss=0.1575, simple_loss=0.2335, pruned_loss=0.04075, over 1433560.19 frames. ], batch size: 47, lr: 9.19e-03, grad_scale: 8.0 +2023-03-21 00:32:28,889 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48980.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:32:39,139 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 00:32:44,978 INFO [train.py:901] (1/2) Epoch 18, batch 1000, loss[loss=0.1387, simple_loss=0.2009, pruned_loss=0.03824, over 7166.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2335, pruned_loss=0.04086, over 1435518.12 frames. ], batch size: 39, lr: 9.19e-03, grad_scale: 8.0 +2023-03-21 00:32:51,046 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.043e+02 2.421e+02 2.913e+02 5.514e+02, threshold=4.842e+02, percent-clipped=1.0 +2023-03-21 00:32:54,613 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=49028.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:33:00,085 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 00:33:07,131 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49053.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:33:10,128 INFO [train.py:901] (1/2) Epoch 18, batch 1050, loss[loss=0.1515, simple_loss=0.228, pruned_loss=0.03744, over 7276.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2337, pruned_loss=0.04068, over 1438204.66 frames. ], batch size: 52, lr: 9.18e-03, grad_scale: 8.0 +2023-03-21 00:33:21,169 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49079.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:33:23,122 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 00:33:27,821 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 00:33:32,934 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=49101.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:33:35,696 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 +2023-03-21 00:33:36,868 INFO [train.py:901] (1/2) Epoch 18, batch 1100, loss[loss=0.1273, simple_loss=0.1913, pruned_loss=0.03169, over 6447.00 frames. ], tot_loss[loss=0.1577, simple_loss=0.2335, pruned_loss=0.0409, over 1436925.24 frames. ], batch size: 28, lr: 9.18e-03, grad_scale: 8.0 +2023-03-21 00:33:42,797 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.227e+02 2.061e+02 2.375e+02 2.807e+02 6.437e+02, threshold=4.750e+02, percent-clipped=1.0 +2023-03-21 00:33:45,901 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=49127.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:33:56,396 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 00:33:56,903 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 00:33:56,974 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.9415, 4.4860, 4.4070, 4.9332, 4.9096, 4.9537, 4.3691, 4.6117], + device='cuda:1'), covar=tensor([0.0841, 0.2450, 0.2229, 0.1124, 0.0797, 0.1316, 0.0759, 0.0972], + device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0326, 0.0258, 0.0258, 0.0187, 0.0322, 0.0184, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:34:01,901 INFO [train.py:901] (1/2) Epoch 18, batch 1150, loss[loss=0.1566, simple_loss=0.2292, pruned_loss=0.04197, over 7320.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2336, pruned_loss=0.04053, over 1439290.44 frames. ], batch size: 75, lr: 9.18e-03, grad_scale: 8.0 +2023-03-21 00:34:08,696 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 00:34:09,702 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 00:34:12,323 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49177.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:34:20,950 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49194.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:34:28,729 INFO [train.py:901] (1/2) Epoch 18, batch 1200, loss[loss=0.1672, simple_loss=0.246, pruned_loss=0.04418, over 7324.00 frames. ], tot_loss[loss=0.1571, simple_loss=0.2333, pruned_loss=0.04043, over 1439787.01 frames. ], batch size: 75, lr: 9.17e-03, grad_scale: 8.0 +2023-03-21 00:34:29,986 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6940, 2.6989, 2.0014, 3.2939, 2.3201, 2.6803, 1.5475, 1.8281], + device='cuda:1'), covar=tensor([0.0310, 0.0606, 0.2292, 0.0605, 0.0427, 0.0382, 0.3340, 0.1686], + device='cuda:1'), in_proj_covar=tensor([0.0259, 0.0241, 0.0301, 0.0248, 0.0258, 0.0255, 0.0264, 0.0279], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 00:34:34,778 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.271e+02 2.292e+02 2.732e+02 3.185e+02 8.511e+02, threshold=5.463e+02, percent-clipped=4.0 +2023-03-21 00:34:42,370 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 00:34:42,502 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49236.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:34:49,134 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0635, 2.8915, 3.2561, 3.0518, 3.1355, 2.9963, 2.4890, 2.9257], + device='cuda:1'), covar=tensor([0.1827, 0.0906, 0.1149, 0.1575, 0.1270, 0.1347, 0.3170, 0.2486], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0049, 0.0038, 0.0039, 0.0037, 0.0035, 0.0053, 0.0041], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 00:34:54,719 INFO [train.py:901] (1/2) Epoch 18, batch 1250, loss[loss=0.1646, simple_loss=0.2404, pruned_loss=0.04435, over 7301.00 frames. ], tot_loss[loss=0.1563, simple_loss=0.2326, pruned_loss=0.04006, over 1437604.68 frames. ], batch size: 86, lr: 9.17e-03, grad_scale: 16.0 +2023-03-21 00:34:58,193 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-03-21 00:35:03,600 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2040, 2.3710, 2.1229, 3.4326, 1.5074, 3.2983, 1.2747, 2.9205], + device='cuda:1'), covar=tensor([0.0076, 0.0949, 0.1624, 0.0082, 0.3708, 0.0141, 0.1068, 0.0283], + device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0261, 0.0290, 0.0172, 0.0274, 0.0188, 0.0260, 0.0219], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 00:35:07,019 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 00:35:11,029 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 00:35:12,018 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 00:35:14,639 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49297.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:35:17,673 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4988, 3.3624, 2.3654, 4.0986, 2.9878, 3.4919, 1.9098, 2.2098], + device='cuda:1'), covar=tensor([0.0357, 0.1093, 0.2197, 0.0432, 0.0459, 0.0599, 0.2825, 0.1755], + device='cuda:1'), in_proj_covar=tensor([0.0259, 0.0240, 0.0300, 0.0247, 0.0257, 0.0254, 0.0261, 0.0278], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 00:35:20,484 INFO [train.py:901] (1/2) Epoch 18, batch 1300, loss[loss=0.1348, simple_loss=0.1957, pruned_loss=0.0369, over 7013.00 frames. ], tot_loss[loss=0.1553, simple_loss=0.2315, pruned_loss=0.03955, over 1438171.15 frames. ], batch size: 35, lr: 9.16e-03, grad_scale: 16.0 +2023-03-21 00:35:26,587 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.386e+02 2.121e+02 2.615e+02 3.045e+02 6.610e+02, threshold=5.231e+02, percent-clipped=1.0 +2023-03-21 00:35:35,211 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 00:35:38,303 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 00:35:42,460 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 00:35:47,081 INFO [train.py:901] (1/2) Epoch 18, batch 1350, loss[loss=0.1438, simple_loss=0.2257, pruned_loss=0.03089, over 7247.00 frames. ], tot_loss[loss=0.156, simple_loss=0.2319, pruned_loss=0.04011, over 1437587.82 frames. ], batch size: 89, lr: 9.16e-03, grad_scale: 16.0 +2023-03-21 00:35:52,566 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 00:35:55,665 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8132, 4.3352, 4.3923, 4.3057, 4.3213, 3.9664, 4.3949, 4.2504], + device='cuda:1'), covar=tensor([0.0510, 0.0444, 0.0404, 0.0504, 0.0358, 0.0377, 0.0351, 0.0534], + device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0209, 0.0148, 0.0154, 0.0125, 0.0192, 0.0162, 0.0125], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:36:12,418 INFO [train.py:901] (1/2) Epoch 18, batch 1400, loss[loss=0.1553, simple_loss=0.2322, pruned_loss=0.0392, over 7303.00 frames. ], tot_loss[loss=0.1564, simple_loss=0.2321, pruned_loss=0.04034, over 1437644.30 frames. ], batch size: 80, lr: 9.15e-03, grad_scale: 16.0 +2023-03-21 00:36:17,241 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-03-21 00:36:18,329 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.396e+02 2.167e+02 2.498e+02 3.013e+02 7.794e+02, threshold=4.996e+02, percent-clipped=2.0 +2023-03-21 00:36:18,934 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49422.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:36:24,469 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 00:36:24,576 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49432.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 00:36:38,495 INFO [train.py:901] (1/2) Epoch 18, batch 1450, loss[loss=0.1596, simple_loss=0.2427, pruned_loss=0.03826, over 7299.00 frames. ], tot_loss[loss=0.1568, simple_loss=0.2327, pruned_loss=0.04047, over 1439551.73 frames. ], batch size: 68, lr: 9.15e-03, grad_scale: 16.0 +2023-03-21 00:36:47,831 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49477.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:36:49,280 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 00:36:50,928 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49483.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:36:55,975 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49493.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 00:36:56,409 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49494.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:37:03,820 INFO [train.py:901] (1/2) Epoch 18, batch 1500, loss[loss=0.1817, simple_loss=0.2483, pruned_loss=0.05754, over 7269.00 frames. ], tot_loss[loss=0.1566, simple_loss=0.2325, pruned_loss=0.04031, over 1439170.88 frames. ], batch size: 47, lr: 9.14e-03, grad_scale: 16.0 +2023-03-21 00:37:05,416 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 00:37:10,441 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.480e+02 2.142e+02 2.435e+02 2.862e+02 6.987e+02, threshold=4.869e+02, percent-clipped=2.0 +2023-03-21 00:37:12,579 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=49525.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:37:13,837 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.35 vs. limit=5.0 +2023-03-21 00:37:20,938 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=49542.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:37:25,291 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.03 vs. limit=5.0 +2023-03-21 00:37:29,285 INFO [train.py:901] (1/2) Epoch 18, batch 1550, loss[loss=0.1525, simple_loss=0.2289, pruned_loss=0.03801, over 7200.00 frames. ], tot_loss[loss=0.1563, simple_loss=0.2324, pruned_loss=0.04004, over 1441014.29 frames. ], batch size: 50, lr: 9.14e-03, grad_scale: 16.0 +2023-03-21 00:37:29,820 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 00:37:37,599 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1555, 1.4763, 1.2599, 1.3689, 1.3751, 1.2005, 1.3127, 1.0322], + device='cuda:1'), covar=tensor([0.0078, 0.0060, 0.0160, 0.0051, 0.0050, 0.0067, 0.0092, 0.0095], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0023, 0.0023, 0.0022, 0.0022, 0.0022, 0.0024, 0.0030], + device='cuda:1'), out_proj_covar=tensor([2.8441e-05, 2.6555e-05, 2.6494e-05, 2.5285e-05, 2.6703e-05, 2.5158e-05, + 2.7727e-05, 3.4886e-05], device='cuda:1') +2023-03-21 00:37:45,877 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49592.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:37:49,732 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-03-21 00:37:55,994 INFO [train.py:901] (1/2) Epoch 18, batch 1600, loss[loss=0.1578, simple_loss=0.225, pruned_loss=0.04528, over 7250.00 frames. ], tot_loss[loss=0.1563, simple_loss=0.2324, pruned_loss=0.04016, over 1441593.87 frames. ], batch size: 45, lr: 9.13e-03, grad_scale: 16.0 +2023-03-21 00:38:02,087 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.138e+02 2.464e+02 3.002e+02 4.958e+02, threshold=4.928e+02, percent-clipped=1.0 +2023-03-21 00:38:02,598 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 00:38:03,623 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 00:38:06,625 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 00:38:15,512 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 00:38:19,480 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 00:38:20,991 INFO [train.py:901] (1/2) Epoch 18, batch 1650, loss[loss=0.1209, simple_loss=0.1945, pruned_loss=0.02358, over 7011.00 frames. ], tot_loss[loss=0.1561, simple_loss=0.2322, pruned_loss=0.03997, over 1441670.21 frames. ], batch size: 35, lr: 9.13e-03, grad_scale: 16.0 +2023-03-21 00:38:25,557 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7551, 3.3256, 3.4997, 3.4507, 3.3352, 3.2728, 3.5693, 3.3426], + device='cuda:1'), covar=tensor([0.0126, 0.0223, 0.0132, 0.0174, 0.0425, 0.0136, 0.0170, 0.0145], + device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0079, 0.0078, 0.0068, 0.0138, 0.0090, 0.0082, 0.0086], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:38:28,015 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 00:38:29,626 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1061, 1.4330, 1.1457, 1.2924, 1.2350, 1.2397, 1.2991, 0.9817], + device='cuda:1'), covar=tensor([0.0134, 0.0089, 0.0162, 0.0059, 0.0093, 0.0061, 0.0118, 0.0125], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0023, 0.0022, 0.0022, 0.0022, 0.0022, 0.0024, 0.0029], + device='cuda:1'), out_proj_covar=tensor([2.7808e-05, 2.6080e-05, 2.6130e-05, 2.4608e-05, 2.6132e-05, 2.4560e-05, + 2.6991e-05, 3.3957e-05], device='cuda:1') +2023-03-21 00:38:30,644 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3150, 1.6033, 1.3269, 1.4394, 1.4821, 1.4752, 1.4887, 1.0745], + device='cuda:1'), covar=tensor([0.0134, 0.0100, 0.0163, 0.0069, 0.0134, 0.0054, 0.0096, 0.0119], + device='cuda:1'), in_proj_covar=tensor([0.0024, 0.0023, 0.0022, 0.0022, 0.0022, 0.0022, 0.0024, 0.0029], + device='cuda:1'), out_proj_covar=tensor([2.7813e-05, 2.6076e-05, 2.6127e-05, 2.4601e-05, 2.6144e-05, 2.4574e-05, + 2.6996e-05, 3.3963e-05], device='cuda:1') +2023-03-21 00:38:45,912 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 00:38:47,428 INFO [train.py:901] (1/2) Epoch 18, batch 1700, loss[loss=0.132, simple_loss=0.2118, pruned_loss=0.02605, over 7095.00 frames. ], tot_loss[loss=0.1561, simple_loss=0.232, pruned_loss=0.04005, over 1438716.90 frames. ], batch size: 41, lr: 9.13e-03, grad_scale: 16.0 +2023-03-21 00:38:49,017 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4813, 4.6980, 4.4994, 4.6554, 4.2922, 4.5535, 4.8849, 4.9454], + device='cuda:1'), covar=tensor([0.0193, 0.0116, 0.0135, 0.0115, 0.0276, 0.0170, 0.0215, 0.0140], + device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0110, 0.0101, 0.0110, 0.0103, 0.0094, 0.0089, 0.0085], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:38:49,937 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 00:38:53,488 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.121e+02 2.517e+02 3.238e+02 5.019e+02, threshold=5.033e+02, percent-clipped=1.0 +2023-03-21 00:39:00,530 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 00:39:08,235 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6780, 2.3816, 2.6997, 2.5099, 2.7854, 2.5068, 2.1816, 2.7357], + device='cuda:1'), covar=tensor([0.1843, 0.0769, 0.1285, 0.1636, 0.0790, 0.1282, 0.2140, 0.1517], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0050, 0.0039, 0.0041, 0.0038, 0.0037, 0.0055, 0.0041], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 00:39:12,631 INFO [train.py:901] (1/2) Epoch 18, batch 1750, loss[loss=0.1577, simple_loss=0.2394, pruned_loss=0.03805, over 7322.00 frames. ], tot_loss[loss=0.1561, simple_loss=0.2323, pruned_loss=0.03998, over 1441775.85 frames. ], batch size: 75, lr: 9.12e-03, grad_scale: 16.0 +2023-03-21 00:39:16,341 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49766.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:39:18,988 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-03-21 00:39:22,282 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49778.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:39:25,565 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1665, 3.6305, 3.8884, 3.8734, 3.7843, 3.7133, 3.9257, 3.5769], + device='cuda:1'), covar=tensor([0.0107, 0.0181, 0.0105, 0.0121, 0.0361, 0.0127, 0.0155, 0.0142], + device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0078, 0.0077, 0.0067, 0.0136, 0.0089, 0.0082, 0.0085], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:39:25,989 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 00:39:26,463 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 00:39:28,535 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49788.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 00:39:32,618 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9613, 3.3128, 2.7320, 2.8873, 3.0453, 2.5525, 3.0491, 2.8992], + device='cuda:1'), covar=tensor([0.0601, 0.0721, 0.1117, 0.1200, 0.1482, 0.1019, 0.0863, 0.0872], + device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0044, 0.0053, 0.0046, 0.0044, 0.0047, 0.0048, 0.0043], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:39:39,293 INFO [train.py:901] (1/2) Epoch 18, batch 1800, loss[loss=0.148, simple_loss=0.2276, pruned_loss=0.03423, over 7309.00 frames. ], tot_loss[loss=0.1561, simple_loss=0.2324, pruned_loss=0.0399, over 1442672.74 frames. ], batch size: 49, lr: 9.12e-03, grad_scale: 16.0 +2023-03-21 00:39:45,279 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.430e+02 2.060e+02 2.380e+02 2.875e+02 7.454e+02, threshold=4.761e+02, percent-clipped=2.0 +2023-03-21 00:39:48,288 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 00:39:48,423 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49827.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:40:00,725 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-03-21 00:40:01,426 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 00:40:04,435 INFO [train.py:901] (1/2) Epoch 18, batch 1850, loss[loss=0.1472, simple_loss=0.2279, pruned_loss=0.03324, over 7211.00 frames. ], tot_loss[loss=0.1557, simple_loss=0.2317, pruned_loss=0.03982, over 1441619.97 frames. ], batch size: 50, lr: 9.11e-03, grad_scale: 16.0 +2023-03-21 00:40:12,180 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 00:40:22,403 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49892.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:40:28,791 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 00:40:30,769 INFO [train.py:901] (1/2) Epoch 18, batch 1900, loss[loss=0.1706, simple_loss=0.2454, pruned_loss=0.04789, over 7307.00 frames. ], tot_loss[loss=0.1554, simple_loss=0.2316, pruned_loss=0.03959, over 1442641.97 frames. ], batch size: 80, lr: 9.11e-03, grad_scale: 16.0 +2023-03-21 00:40:36,809 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.351e+02 2.153e+02 2.525e+02 3.111e+02 5.745e+02, threshold=5.051e+02, percent-clipped=1.0 +2023-03-21 00:40:46,607 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=49940.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:40:53,665 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 00:40:57,235 INFO [train.py:901] (1/2) Epoch 18, batch 1950, loss[loss=0.1569, simple_loss=0.2396, pruned_loss=0.03711, over 7258.00 frames. ], tot_loss[loss=0.1553, simple_loss=0.2314, pruned_loss=0.03957, over 1442565.57 frames. ], batch size: 89, lr: 9.10e-03, grad_scale: 16.0 +2023-03-21 00:41:05,306 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 00:41:10,250 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 00:41:10,773 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 00:41:22,513 INFO [train.py:901] (1/2) Epoch 18, batch 2000, loss[loss=0.1545, simple_loss=0.2291, pruned_loss=0.03996, over 7314.00 frames. ], tot_loss[loss=0.1564, simple_loss=0.2324, pruned_loss=0.04021, over 1443366.90 frames. ], batch size: 75, lr: 9.10e-03, grad_scale: 16.0 +2023-03-21 00:41:27,150 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 00:41:28,250 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0145, 3.5315, 4.0949, 4.0643, 4.1118, 4.0990, 4.0740, 3.8147], + device='cuda:1'), covar=tensor([0.0044, 0.0129, 0.0050, 0.0042, 0.0039, 0.0039, 0.0047, 0.0069], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0054, 0.0047, 0.0045, 0.0046, 0.0049, 0.0047, 0.0059], + device='cuda:1'), out_proj_covar=tensor([8.4277e-05, 1.3398e-04, 1.1029e-04, 9.7850e-05, 1.0009e-04, 1.0488e-04, + 1.1324e-04, 1.3125e-04], device='cuda:1') +2023-03-21 00:41:28,631 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.390e+02 2.229e+02 2.681e+02 3.198e+02 6.835e+02, threshold=5.363e+02, percent-clipped=3.0 +2023-03-21 00:41:37,827 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0737, 2.8206, 2.9043, 2.9837, 3.2080, 2.9489, 2.5818, 3.2098], + device='cuda:1'), covar=tensor([0.1736, 0.0562, 0.1228, 0.1234, 0.0615, 0.1222, 0.2244, 0.1206], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0051, 0.0040, 0.0040, 0.0039, 0.0037, 0.0055, 0.0041], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 00:41:39,237 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 00:41:46,992 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5437, 4.9680, 5.1040, 5.0106, 4.8715, 4.5745, 5.0862, 4.9433], + device='cuda:1'), covar=tensor([0.0400, 0.0413, 0.0346, 0.0430, 0.0333, 0.0320, 0.0354, 0.0472], + device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0213, 0.0153, 0.0158, 0.0128, 0.0196, 0.0165, 0.0127], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:41:47,439 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 00:41:48,935 INFO [train.py:901] (1/2) Epoch 18, batch 2050, loss[loss=0.1613, simple_loss=0.2412, pruned_loss=0.04072, over 7332.00 frames. ], tot_loss[loss=0.1562, simple_loss=0.2322, pruned_loss=0.04015, over 1442678.41 frames. ], batch size: 61, lr: 9.09e-03, grad_scale: 16.0 +2023-03-21 00:41:58,472 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50078.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:42:01,394 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.3887, 4.9328, 4.9849, 5.3330, 5.3811, 5.3257, 4.8228, 4.9935], + device='cuda:1'), covar=tensor([0.0676, 0.2180, 0.1812, 0.1009, 0.0655, 0.1188, 0.0563, 0.0902], + device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0325, 0.0253, 0.0257, 0.0188, 0.0318, 0.0182, 0.0231], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:42:03,482 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50088.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 00:42:13,955 INFO [train.py:901] (1/2) Epoch 18, batch 2100, loss[loss=0.1618, simple_loss=0.2419, pruned_loss=0.04083, over 7326.00 frames. ], tot_loss[loss=0.1564, simple_loss=0.2324, pruned_loss=0.04026, over 1440576.03 frames. ], batch size: 75, lr: 9.09e-03, grad_scale: 16.0 +2023-03-21 00:42:20,555 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.448e+02 1.971e+02 2.438e+02 2.876e+02 7.861e+02, threshold=4.875e+02, percent-clipped=1.0 +2023-03-21 00:42:20,575 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 00:42:21,151 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50122.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:42:23,111 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 00:42:23,149 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=50126.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:42:24,301 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50128.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:42:28,863 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=50136.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 00:42:40,326 INFO [train.py:901] (1/2) Epoch 18, batch 2150, loss[loss=0.139, simple_loss=0.2158, pruned_loss=0.03114, over 7331.00 frames. ], tot_loss[loss=0.1565, simple_loss=0.2325, pruned_loss=0.04025, over 1440386.58 frames. ], batch size: 44, lr: 9.08e-03, grad_scale: 16.0 +2023-03-21 00:42:55,381 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50189.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:42:59,312 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50197.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 00:43:06,079 INFO [train.py:901] (1/2) Epoch 18, batch 2200, loss[loss=0.1464, simple_loss=0.2214, pruned_loss=0.03572, over 7134.00 frames. ], tot_loss[loss=0.1574, simple_loss=0.2331, pruned_loss=0.04083, over 1441653.67 frames. ], batch size: 41, lr: 9.08e-03, grad_scale: 16.0 +2023-03-21 00:43:08,553 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 00:43:12,640 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.073e+02 2.473e+02 2.866e+02 5.110e+02, threshold=4.946e+02, percent-clipped=1.0 +2023-03-21 00:43:13,275 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5630, 2.4563, 2.6052, 2.5803, 2.6171, 2.5821, 2.1788, 2.7565], + device='cuda:1'), covar=tensor([0.2400, 0.0839, 0.1293, 0.1482, 0.1168, 0.1514, 0.2577, 0.1393], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0052, 0.0040, 0.0042, 0.0040, 0.0038, 0.0056, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 00:43:31,407 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50258.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 00:43:31,732 INFO [train.py:901] (1/2) Epoch 18, batch 2250, loss[loss=0.16, simple_loss=0.2393, pruned_loss=0.04039, over 7309.00 frames. ], tot_loss[loss=0.1567, simple_loss=0.2327, pruned_loss=0.04034, over 1441367.56 frames. ], batch size: 86, lr: 9.08e-03, grad_scale: 16.0 +2023-03-21 00:43:43,972 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 00:43:43,983 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 00:43:58,194 INFO [train.py:901] (1/2) Epoch 18, batch 2300, loss[loss=0.1529, simple_loss=0.2335, pruned_loss=0.03612, over 7348.00 frames. ], tot_loss[loss=0.1562, simple_loss=0.232, pruned_loss=0.04017, over 1438864.85 frames. ], batch size: 73, lr: 9.07e-03, grad_scale: 16.0 +2023-03-21 00:43:58,216 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 00:44:04,184 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.450e+02 2.045e+02 2.583e+02 3.047e+02 4.891e+02, threshold=5.165e+02, percent-clipped=0.0 +2023-03-21 00:44:23,285 INFO [train.py:901] (1/2) Epoch 18, batch 2350, loss[loss=0.1541, simple_loss=0.2259, pruned_loss=0.04113, over 7333.00 frames. ], tot_loss[loss=0.1559, simple_loss=0.2319, pruned_loss=0.03995, over 1440885.26 frames. ], batch size: 51, lr: 9.07e-03, grad_scale: 8.0 +2023-03-21 00:44:30,035 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9803, 2.5019, 2.0389, 2.7644, 2.5213, 2.5569, 2.6685, 2.3493], + device='cuda:1'), covar=tensor([0.1985, 0.0795, 0.3140, 0.0531, 0.0128, 0.0082, 0.0271, 0.0251], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0233, 0.0268, 0.0263, 0.0154, 0.0148, 0.0181, 0.0197], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:44:45,982 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 00:44:50,017 INFO [train.py:901] (1/2) Epoch 18, batch 2400, loss[loss=0.1707, simple_loss=0.2493, pruned_loss=0.04605, over 7266.00 frames. ], tot_loss[loss=0.1561, simple_loss=0.2322, pruned_loss=0.03998, over 1443964.91 frames. ], batch size: 89, lr: 9.06e-03, grad_scale: 8.0 +2023-03-21 00:44:52,562 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 00:44:56,600 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.401e+02 1.959e+02 2.320e+02 2.795e+02 6.835e+02, threshold=4.641e+02, percent-clipped=1.0 +2023-03-21 00:44:56,725 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50422.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:44:57,723 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5698, 3.9797, 4.1832, 4.1912, 4.1581, 4.1420, 4.4814, 3.9087], + device='cuda:1'), covar=tensor([0.0112, 0.0139, 0.0115, 0.0135, 0.0356, 0.0086, 0.0115, 0.0164], + device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0078, 0.0077, 0.0070, 0.0137, 0.0089, 0.0082, 0.0086], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:45:03,282 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 00:45:05,823 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 00:45:11,908 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2366, 3.7841, 3.8814, 3.9149, 3.7775, 3.8000, 4.1233, 3.5722], + device='cuda:1'), covar=tensor([0.0146, 0.0135, 0.0133, 0.0140, 0.0427, 0.0101, 0.0142, 0.0181], + device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0077, 0.0077, 0.0069, 0.0136, 0.0088, 0.0081, 0.0085], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:45:15,321 INFO [train.py:901] (1/2) Epoch 18, batch 2450, loss[loss=0.164, simple_loss=0.2345, pruned_loss=0.04674, over 7261.00 frames. ], tot_loss[loss=0.1554, simple_loss=0.2318, pruned_loss=0.03948, over 1446613.59 frames. ], batch size: 47, lr: 9.06e-03, grad_scale: 8.0 +2023-03-21 00:45:21,417 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=50470.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:45:28,424 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50484.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:45:31,386 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 +2023-03-21 00:45:32,919 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 00:45:41,444 INFO [train.py:901] (1/2) Epoch 18, batch 2500, loss[loss=0.1556, simple_loss=0.2327, pruned_loss=0.03921, over 7327.00 frames. ], tot_loss[loss=0.1557, simple_loss=0.2322, pruned_loss=0.03965, over 1446360.47 frames. ], batch size: 83, lr: 9.05e-03, grad_scale: 8.0 +2023-03-21 00:45:47,923 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 1.929e+02 2.329e+02 2.819e+02 6.164e+02, threshold=4.658e+02, percent-clipped=3.0 +2023-03-21 00:45:51,108 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9735, 2.2894, 2.0461, 2.7346, 2.5921, 2.7579, 2.6950, 2.3676], + device='cuda:1'), covar=tensor([0.1842, 0.0839, 0.3094, 0.0544, 0.0115, 0.0064, 0.0260, 0.0251], + device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0235, 0.0271, 0.0262, 0.0154, 0.0149, 0.0184, 0.0201], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:45:55,895 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-21 00:45:58,163 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 00:46:00,239 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50546.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:46:01,789 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9650, 2.9768, 2.3912, 3.9197, 1.5031, 3.6038, 1.3583, 2.9579], + device='cuda:1'), covar=tensor([0.0105, 0.0781, 0.1591, 0.0107, 0.4348, 0.0174, 0.1341, 0.0418], + device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0256, 0.0281, 0.0171, 0.0268, 0.0185, 0.0253, 0.0216], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 00:46:03,690 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50553.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 00:46:07,300 INFO [train.py:901] (1/2) Epoch 18, batch 2550, loss[loss=0.1611, simple_loss=0.2292, pruned_loss=0.04644, over 7266.00 frames. ], tot_loss[loss=0.1548, simple_loss=0.231, pruned_loss=0.03931, over 1444353.69 frames. ], batch size: 47, lr: 9.05e-03, grad_scale: 8.0 +2023-03-21 00:46:32,561 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50607.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:46:33,443 INFO [train.py:901] (1/2) Epoch 18, batch 2600, loss[loss=0.1579, simple_loss=0.2258, pruned_loss=0.04494, over 7311.00 frames. ], tot_loss[loss=0.155, simple_loss=0.2315, pruned_loss=0.03922, over 1444865.74 frames. ], batch size: 49, lr: 9.04e-03, grad_scale: 8.0 +2023-03-21 00:46:36,120 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7379, 2.8634, 1.9380, 3.5062, 2.0590, 2.7839, 1.4743, 1.9071], + device='cuda:1'), covar=tensor([0.0330, 0.1063, 0.2426, 0.0527, 0.0376, 0.0630, 0.3154, 0.1739], + device='cuda:1'), in_proj_covar=tensor([0.0260, 0.0242, 0.0294, 0.0249, 0.0260, 0.0254, 0.0261, 0.0276], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 00:46:39,854 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.366e+02 2.102e+02 2.428e+02 3.097e+02 8.694e+02, threshold=4.856e+02, percent-clipped=5.0 +2023-03-21 00:46:58,170 INFO [train.py:901] (1/2) Epoch 18, batch 2650, loss[loss=0.1946, simple_loss=0.2694, pruned_loss=0.05987, over 7150.00 frames. ], tot_loss[loss=0.1551, simple_loss=0.2316, pruned_loss=0.03933, over 1442528.14 frames. ], batch size: 98, lr: 9.04e-03, grad_scale: 8.0 +2023-03-21 00:47:22,875 INFO [train.py:901] (1/2) Epoch 18, batch 2700, loss[loss=0.1705, simple_loss=0.2467, pruned_loss=0.04722, over 7294.00 frames. ], tot_loss[loss=0.1552, simple_loss=0.2319, pruned_loss=0.03927, over 1441999.22 frames. ], batch size: 80, lr: 9.04e-03, grad_scale: 8.0 +2023-03-21 00:47:29,654 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.009e+02 2.432e+02 3.028e+02 8.989e+02, threshold=4.865e+02, percent-clipped=2.0 +2023-03-21 00:47:31,230 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2573, 4.7567, 4.7991, 4.7652, 4.6383, 4.3359, 4.8343, 4.6814], + device='cuda:1'), covar=tensor([0.0475, 0.0459, 0.0412, 0.0462, 0.0404, 0.0361, 0.0350, 0.0519], + device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0217, 0.0157, 0.0158, 0.0132, 0.0199, 0.0169, 0.0130], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:47:47,625 INFO [train.py:901] (1/2) Epoch 18, batch 2750, loss[loss=0.1392, simple_loss=0.2217, pruned_loss=0.02832, over 7354.00 frames. ], tot_loss[loss=0.1557, simple_loss=0.2322, pruned_loss=0.0396, over 1444424.04 frames. ], batch size: 63, lr: 9.03e-03, grad_scale: 8.0 +2023-03-21 00:47:59,188 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50782.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:47:59,850 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-21 00:48:00,554 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50784.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:48:06,793 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9949, 2.0907, 2.4222, 1.9008, 1.9250, 2.1614, 1.9566, 1.5229], + device='cuda:1'), covar=tensor([0.0349, 0.0420, 0.0140, 0.0088, 0.0357, 0.0239, 0.0159, 0.0238], + device='cuda:1'), in_proj_covar=tensor([0.0026, 0.0026, 0.0023, 0.0022, 0.0025, 0.0023, 0.0027, 0.0026], + device='cuda:1'), out_proj_covar=tensor([6.5962e-05, 6.5423e-05, 5.7602e-05, 5.5607e-05, 6.2903e-05, 5.9147e-05, + 6.4823e-05, 6.6594e-05], device='cuda:1') +2023-03-21 00:48:12,417 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50808.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 00:48:12,740 INFO [train.py:901] (1/2) Epoch 18, batch 2800, loss[loss=0.1568, simple_loss=0.2378, pruned_loss=0.0379, over 7302.00 frames. ], tot_loss[loss=0.1564, simple_loss=0.2328, pruned_loss=0.03999, over 1443139.70 frames. ], batch size: 68, lr: 9.03e-03, grad_scale: 8.0 +2023-03-21 00:48:17,145 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2643, 4.7925, 4.8957, 4.7888, 4.6885, 4.4100, 4.9167, 4.7098], + device='cuda:1'), covar=tensor([0.0420, 0.0394, 0.0340, 0.0398, 0.0316, 0.0327, 0.0292, 0.0489], + device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0214, 0.0154, 0.0154, 0.0129, 0.0196, 0.0165, 0.0127], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:48:18,918 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.308e+02 1.954e+02 2.260e+02 2.876e+02 5.616e+02, threshold=4.520e+02, percent-clipped=4.0 +2023-03-21 00:48:37,699 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 00:48:37,752 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 00:48:37,775 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 00:48:37,793 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 00:48:37,963 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 00:48:37,977 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 00:48:37,989 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 00:48:38,089 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 00:48:38,163 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 00:48:38,177 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 00:48:38,194 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 00:48:38,680 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 00:48:38,686 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 00:48:38,694 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 00:48:38,829 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 00:48:38,832 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 00:48:38,889 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 00:48:45,778 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=50832.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:48:46,223 INFO [train.py:901] (1/2) Epoch 19, batch 0, loss[loss=0.1525, simple_loss=0.234, pruned_loss=0.03549, over 7261.00 frames. ], tot_loss[loss=0.1525, simple_loss=0.234, pruned_loss=0.03549, over 7261.00 frames. ], batch size: 89, lr: 8.80e-03, grad_scale: 8.0 +2023-03-21 00:48:46,223 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 00:49:03,775 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.8650, 5.1121, 5.1989, 5.1789, 4.7936, 4.7836, 5.2383, 4.8330], + device='cuda:1'), covar=tensor([0.0327, 0.0366, 0.0296, 0.0366, 0.0406, 0.0274, 0.0285, 0.0596], + device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0213, 0.0153, 0.0154, 0.0129, 0.0195, 0.0164, 0.0127], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:49:10,800 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7739, 4.5486, 4.2360, 4.9809, 4.7735, 5.0180, 4.5681, 4.7583], + device='cuda:1'), covar=tensor([0.0832, 0.1824, 0.2194, 0.1275, 0.0859, 0.0872, 0.0508, 0.0655], + device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0327, 0.0257, 0.0255, 0.0188, 0.0318, 0.0186, 0.0232], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:49:12,200 INFO [train.py:935] (1/2) Epoch 19, validation: loss=0.166, simple_loss=0.2533, pruned_loss=0.03935, over 1622729.00 frames. +2023-03-21 00:49:12,200 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 00:49:17,283 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50843.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:49:18,728 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 00:49:22,181 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50853.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 00:49:30,082 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 00:49:30,203 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50869.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 00:49:37,072 INFO [train.py:901] (1/2) Epoch 19, batch 50, loss[loss=0.1499, simple_loss=0.2252, pruned_loss=0.03731, over 7161.00 frames. ], tot_loss[loss=0.1565, simple_loss=0.2328, pruned_loss=0.04011, over 327116.31 frames. ], batch size: 41, lr: 8.79e-03, grad_scale: 8.0 +2023-03-21 00:49:37,086 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 00:49:39,146 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 00:49:42,575 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 00:49:46,159 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=50901.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 00:49:46,652 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50902.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:49:47,654 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0502, 4.5338, 4.5933, 4.5608, 4.5863, 4.1808, 4.6715, 4.5425], + device='cuda:1'), covar=tensor([0.0481, 0.0469, 0.0477, 0.0484, 0.0316, 0.0342, 0.0350, 0.0497], + device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0213, 0.0154, 0.0155, 0.0130, 0.0195, 0.0166, 0.0128], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:49:57,769 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.168e+02 2.080e+02 2.542e+02 3.231e+02 6.129e+02, threshold=5.083e+02, percent-clipped=6.0 +2023-03-21 00:50:01,291 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 00:50:02,302 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 00:50:03,348 INFO [train.py:901] (1/2) Epoch 19, batch 100, loss[loss=0.1208, simple_loss=0.1934, pruned_loss=0.02412, over 7032.00 frames. ], tot_loss[loss=0.1566, simple_loss=0.2333, pruned_loss=0.03996, over 574943.31 frames. ], batch size: 35, lr: 8.79e-03, grad_scale: 8.0 +2023-03-21 00:50:10,064 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50946.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:50:13,482 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2611, 4.8299, 4.9204, 4.8304, 4.7263, 4.3582, 4.9556, 4.7344], + device='cuda:1'), covar=tensor([0.0447, 0.0370, 0.0333, 0.0380, 0.0343, 0.0360, 0.0271, 0.0517], + device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0211, 0.0153, 0.0155, 0.0129, 0.0194, 0.0165, 0.0128], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:50:28,296 INFO [train.py:901] (1/2) Epoch 19, batch 150, loss[loss=0.178, simple_loss=0.252, pruned_loss=0.05198, over 7347.00 frames. ], tot_loss[loss=0.1567, simple_loss=0.2332, pruned_loss=0.04009, over 769277.54 frames. ], batch size: 63, lr: 8.78e-03, grad_scale: 8.0 +2023-03-21 00:50:40,554 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3050, 1.3134, 1.2325, 1.3125, 1.4411, 1.2845, 1.4333, 1.1028], + device='cuda:1'), covar=tensor([0.0093, 0.0121, 0.0262, 0.0090, 0.0079, 0.0100, 0.0084, 0.0103], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0024, 0.0023, 0.0024, 0.0023, 0.0023, 0.0025, 0.0030], + device='cuda:1'), out_proj_covar=tensor([2.9366e-05, 2.7428e-05, 2.6929e-05, 2.6523e-05, 2.7089e-05, 2.5934e-05, + 2.8384e-05, 3.5252e-05], device='cuda:1') +2023-03-21 00:50:42,201 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51007.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:50:49,623 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.504e+02 2.085e+02 2.413e+02 2.711e+02 7.121e+02, threshold=4.826e+02, percent-clipped=1.0 +2023-03-21 00:50:55,262 INFO [train.py:901] (1/2) Epoch 19, batch 200, loss[loss=0.1538, simple_loss=0.2345, pruned_loss=0.03656, over 7343.00 frames. ], tot_loss[loss=0.1557, simple_loss=0.2324, pruned_loss=0.03949, over 916153.91 frames. ], batch size: 61, lr: 8.78e-03, grad_scale: 8.0 +2023-03-21 00:51:03,003 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 00:51:07,521 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 00:51:10,134 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51062.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:51:12,501 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 00:51:16,669 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2520, 0.9919, 1.3281, 1.5020, 1.5244, 1.7944, 1.1451, 1.5469], + device='cuda:1'), covar=tensor([0.1130, 0.3423, 0.1349, 0.1095, 0.1568, 0.1774, 0.2379, 0.1168], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0057, 0.0042, 0.0040, 0.0043, 0.0044, 0.0062, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 00:51:21,236 INFO [train.py:901] (1/2) Epoch 19, batch 250, loss[loss=0.146, simple_loss=0.2236, pruned_loss=0.03419, over 7338.00 frames. ], tot_loss[loss=0.1554, simple_loss=0.2319, pruned_loss=0.03945, over 1033130.89 frames. ], batch size: 54, lr: 8.78e-03, grad_scale: 8.0 +2023-03-21 00:51:27,222 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 00:51:41,225 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.398e+02 2.027e+02 2.472e+02 3.001e+02 8.354e+02, threshold=4.945e+02, percent-clipped=6.0 +2023-03-21 00:51:41,911 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51123.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:51:46,652 INFO [train.py:901] (1/2) Epoch 19, batch 300, loss[loss=0.1593, simple_loss=0.2327, pruned_loss=0.04296, over 7249.00 frames. ], tot_loss[loss=0.1548, simple_loss=0.231, pruned_loss=0.03931, over 1122505.10 frames. ], batch size: 55, lr: 8.77e-03, grad_scale: 8.0 +2023-03-21 00:51:48,666 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 00:51:49,204 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51138.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:51:52,977 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.88 vs. limit=5.0 +2023-03-21 00:51:57,106 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 00:52:01,178 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1760, 2.1811, 2.2099, 3.4827, 1.5089, 3.3467, 1.3362, 3.0792], + device='cuda:1'), covar=tensor([0.0093, 0.0880, 0.1326, 0.0105, 0.3483, 0.0148, 0.0992, 0.0209], + device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0259, 0.0286, 0.0175, 0.0274, 0.0190, 0.0258, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 00:52:02,113 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51164.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 00:52:06,759 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51172.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 00:52:12,030 INFO [train.py:901] (1/2) Epoch 19, batch 350, loss[loss=0.1604, simple_loss=0.2409, pruned_loss=0.03993, over 7276.00 frames. ], tot_loss[loss=0.1553, simple_loss=0.2316, pruned_loss=0.03945, over 1194313.86 frames. ], batch size: 57, lr: 8.77e-03, grad_scale: 8.0 +2023-03-21 00:52:22,684 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51202.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:52:26,076 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.3530, 4.9322, 4.7523, 5.3454, 5.1658, 5.3262, 4.6097, 4.9228], + device='cuda:1'), covar=tensor([0.0643, 0.1938, 0.1829, 0.0712, 0.0684, 0.0831, 0.0558, 0.0841], + device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0326, 0.0259, 0.0255, 0.0190, 0.0318, 0.0183, 0.0233], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:52:31,004 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 00:52:32,476 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.486e+02 2.221e+02 2.732e+02 3.384e+02 5.150e+02, threshold=5.463e+02, percent-clipped=2.0 +2023-03-21 00:52:35,183 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1661, 2.2755, 2.1396, 3.5511, 1.5122, 3.4109, 1.3338, 3.2350], + device='cuda:1'), covar=tensor([0.0083, 0.0967, 0.1487, 0.0099, 0.3575, 0.0181, 0.1051, 0.0234], + device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0258, 0.0283, 0.0175, 0.0273, 0.0188, 0.0257, 0.0220], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 00:52:38,066 INFO [train.py:901] (1/2) Epoch 19, batch 400, loss[loss=0.1537, simple_loss=0.2387, pruned_loss=0.03432, over 7149.00 frames. ], tot_loss[loss=0.1551, simple_loss=0.2318, pruned_loss=0.03927, over 1249631.04 frames. ], batch size: 98, lr: 8.76e-03, grad_scale: 8.0 +2023-03-21 00:52:38,212 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51233.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 00:52:46,636 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=51250.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:53:04,731 INFO [train.py:901] (1/2) Epoch 19, batch 450, loss[loss=0.1393, simple_loss=0.2045, pruned_loss=0.03699, over 6958.00 frames. ], tot_loss[loss=0.154, simple_loss=0.2307, pruned_loss=0.03863, over 1292699.16 frames. ], batch size: 35, lr: 8.76e-03, grad_scale: 8.0 +2023-03-21 00:53:09,551 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6073, 5.1834, 5.2241, 5.1865, 4.9244, 4.6761, 5.2656, 5.0499], + device='cuda:1'), covar=tensor([0.0414, 0.0350, 0.0369, 0.0412, 0.0362, 0.0264, 0.0337, 0.0423], + device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0212, 0.0156, 0.0156, 0.0131, 0.0198, 0.0168, 0.0129], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:53:13,533 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 00:53:14,056 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 00:53:14,617 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51302.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:53:16,733 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1349, 2.5789, 3.1259, 3.1140, 3.3206, 2.9255, 2.6308, 3.2624], + device='cuda:1'), covar=tensor([0.1822, 0.0884, 0.1645, 0.0887, 0.0623, 0.1056, 0.2072, 0.0976], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0053, 0.0040, 0.0040, 0.0039, 0.0038, 0.0056, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 00:53:24,707 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.185e+02 2.176e+02 2.553e+02 3.010e+02 4.956e+02, threshold=5.106e+02, percent-clipped=0.0 +2023-03-21 00:53:28,409 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7596, 2.0448, 1.7291, 2.7662, 2.3877, 2.1030, 1.8240, 2.3100], + device='cuda:1'), covar=tensor([0.1758, 0.0820, 0.2939, 0.0741, 0.0132, 0.0114, 0.0201, 0.0261], + device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0232, 0.0269, 0.0261, 0.0154, 0.0151, 0.0185, 0.0198], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:53:28,496 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 +2023-03-21 00:53:30,233 INFO [train.py:901] (1/2) Epoch 19, batch 500, loss[loss=0.1489, simple_loss=0.2285, pruned_loss=0.03467, over 7282.00 frames. ], tot_loss[loss=0.1546, simple_loss=0.2309, pruned_loss=0.03912, over 1323783.83 frames. ], batch size: 66, lr: 8.75e-03, grad_scale: 8.0 +2023-03-21 00:53:45,719 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7030, 3.5995, 3.2113, 3.4503, 2.8259, 2.6073, 3.6853, 2.5345], + device='cuda:1'), covar=tensor([0.0321, 0.0378, 0.0391, 0.0499, 0.0638, 0.0782, 0.0490, 0.1412], + device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0320, 0.0261, 0.0344, 0.0308, 0.0301, 0.0333, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 00:53:46,511 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 00:53:48,022 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 00:53:48,518 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 00:53:50,506 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 00:53:55,356 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 00:53:56,009 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51382.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:53:56,371 INFO [train.py:901] (1/2) Epoch 19, batch 550, loss[loss=0.1512, simple_loss=0.2352, pruned_loss=0.03363, over 7249.00 frames. ], tot_loss[loss=0.1547, simple_loss=0.2311, pruned_loss=0.03911, over 1350518.72 frames. ], batch size: 55, lr: 8.75e-03, grad_scale: 8.0 +2023-03-21 00:54:01,099 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51392.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:54:06,362 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 00:54:13,789 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 00:54:14,356 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51418.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:54:16,260 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.429e+02 2.200e+02 2.639e+02 3.099e+02 6.064e+02, threshold=5.278e+02, percent-clipped=2.0 +2023-03-21 00:54:16,295 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 00:54:22,250 INFO [train.py:901] (1/2) Epoch 19, batch 600, loss[loss=0.1441, simple_loss=0.2251, pruned_loss=0.03152, over 7293.00 frames. ], tot_loss[loss=0.155, simple_loss=0.2316, pruned_loss=0.03919, over 1368984.42 frames. ], batch size: 86, lr: 8.75e-03, grad_scale: 8.0 +2023-03-21 00:54:23,762 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 00:54:24,823 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51438.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:54:27,986 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51443.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:54:32,039 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.95 vs. limit=5.0 +2023-03-21 00:54:32,956 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51453.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:54:34,968 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8917, 2.3517, 1.6942, 2.8564, 2.6081, 2.2222, 2.2656, 2.5068], + device='cuda:1'), covar=tensor([0.1900, 0.0855, 0.3348, 0.0461, 0.0131, 0.0080, 0.0215, 0.0301], + device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0231, 0.0267, 0.0262, 0.0155, 0.0150, 0.0186, 0.0199], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:54:38,118 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-03-21 00:54:38,422 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51464.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 00:54:39,780 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 00:54:41,443 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5822, 2.8526, 2.3943, 2.8426, 2.8212, 2.4920, 2.7438, 2.6625], + device='cuda:1'), covar=tensor([0.1328, 0.1117, 0.2036, 0.1327, 0.1041, 0.0859, 0.1362, 0.1724], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0046, 0.0054, 0.0047, 0.0045, 0.0047, 0.0048, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:54:45,443 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1704, 3.2369, 2.1816, 3.8162, 2.7711, 3.1877, 1.6393, 2.0483], + device='cuda:1'), covar=tensor([0.0408, 0.0767, 0.2079, 0.0593, 0.0407, 0.0497, 0.2852, 0.1766], + device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0238, 0.0293, 0.0252, 0.0262, 0.0251, 0.0258, 0.0272], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 00:54:47,681 INFO [train.py:901] (1/2) Epoch 19, batch 650, loss[loss=0.1569, simple_loss=0.231, pruned_loss=0.04141, over 7244.00 frames. ], tot_loss[loss=0.1542, simple_loss=0.2307, pruned_loss=0.03886, over 1382656.56 frames. ], batch size: 55, lr: 8.74e-03, grad_scale: 8.0 +2023-03-21 00:54:48,752 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 00:54:49,252 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=51486.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:54:52,250 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5653, 2.8320, 2.2967, 2.7483, 2.7700, 2.4125, 2.7869, 2.7056], + device='cuda:1'), covar=tensor([0.1219, 0.0591, 0.1489, 0.1260, 0.0656, 0.0777, 0.0998, 0.0938], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0046, 0.0054, 0.0047, 0.0045, 0.0047, 0.0048, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:54:53,767 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51495.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:55:00,457 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.86 vs. limit=5.0 +2023-03-21 00:55:02,945 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=51512.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 00:55:07,541 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 00:55:07,995 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.287e+02 2.099e+02 2.381e+02 2.975e+02 4.917e+02, threshold=4.762e+02, percent-clipped=0.0 +2023-03-21 00:55:11,189 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51528.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 00:55:14,436 INFO [train.py:901] (1/2) Epoch 19, batch 700, loss[loss=0.1543, simple_loss=0.2338, pruned_loss=0.03743, over 7357.00 frames. ], tot_loss[loss=0.1542, simple_loss=0.2306, pruned_loss=0.03891, over 1396793.18 frames. ], batch size: 63, lr: 8.74e-03, grad_scale: 8.0 +2023-03-21 00:55:16,487 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 00:55:26,081 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51556.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:55:30,152 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-03-21 00:55:39,512 INFO [train.py:901] (1/2) Epoch 19, batch 750, loss[loss=0.1232, simple_loss=0.1919, pruned_loss=0.02729, over 6981.00 frames. ], tot_loss[loss=0.1532, simple_loss=0.2296, pruned_loss=0.03837, over 1404365.77 frames. ], batch size: 35, lr: 8.73e-03, grad_scale: 8.0 +2023-03-21 00:55:41,033 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 00:55:41,535 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 00:55:50,335 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51602.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:55:56,783 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 00:56:00,947 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.391e+02 2.035e+02 2.484e+02 2.955e+02 5.287e+02, threshold=4.968e+02, percent-clipped=1.0 +2023-03-21 00:56:01,510 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 00:56:05,444 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-21 00:56:06,524 INFO [train.py:901] (1/2) Epoch 19, batch 800, loss[loss=0.173, simple_loss=0.2466, pruned_loss=0.04966, over 7262.00 frames. ], tot_loss[loss=0.1539, simple_loss=0.2304, pruned_loss=0.03868, over 1413291.90 frames. ], batch size: 55, lr: 8.73e-03, grad_scale: 8.0 +2023-03-21 00:56:06,552 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 00:56:07,560 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 00:56:08,680 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9291, 2.4677, 2.7823, 2.8110, 3.1061, 2.6911, 2.3791, 2.7004], + device='cuda:1'), covar=tensor([0.1814, 0.1062, 0.1710, 0.2250, 0.0890, 0.1830, 0.2786, 0.2420], + device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0054, 0.0041, 0.0042, 0.0040, 0.0039, 0.0057, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 00:56:15,111 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=51650.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:56:15,386 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 +2023-03-21 00:56:17,541 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 00:56:32,138 INFO [train.py:901] (1/2) Epoch 19, batch 850, loss[loss=0.1528, simple_loss=0.2279, pruned_loss=0.03888, over 7283.00 frames. ], tot_loss[loss=0.1548, simple_loss=0.2317, pruned_loss=0.039, over 1421329.21 frames. ], batch size: 68, lr: 8.73e-03, grad_scale: 8.0 +2023-03-21 00:56:36,169 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 00:56:36,173 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 00:56:41,906 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 00:56:44,981 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 00:56:46,834 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 +2023-03-21 00:56:50,625 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51718.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:56:52,533 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.416e+02 2.319e+02 2.686e+02 3.166e+02 5.097e+02, threshold=5.372e+02, percent-clipped=1.0 +2023-03-21 00:56:57,946 INFO [train.py:901] (1/2) Epoch 19, batch 900, loss[loss=0.1714, simple_loss=0.2481, pruned_loss=0.04736, over 7132.00 frames. ], tot_loss[loss=0.1547, simple_loss=0.2313, pruned_loss=0.039, over 1424292.67 frames. ], batch size: 98, lr: 8.72e-03, grad_scale: 8.0 +2023-03-21 00:57:00,474 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51738.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:57:05,073 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51747.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:57:05,519 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51748.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:57:08,087 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7272, 3.6793, 3.1327, 3.1231, 2.8921, 2.1546, 1.4899, 3.7908], + device='cuda:1'), covar=tensor([0.0039, 0.0060, 0.0115, 0.0069, 0.0126, 0.0414, 0.0524, 0.0043], + device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0073, 0.0092, 0.0081, 0.0101, 0.0117, 0.0118, 0.0086], + device='cuda:1'), out_proj_covar=tensor([1.0490e-04, 9.8985e-05, 1.1757e-04, 1.0765e-04, 1.2751e-04, 1.4890e-04, + 1.5137e-04, 1.0702e-04], device='cuda:1') +2023-03-21 00:57:11,135 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8520, 2.4097, 1.7282, 2.8294, 2.7147, 2.5837, 2.5083, 2.4980], + device='cuda:1'), covar=tensor([0.1950, 0.0852, 0.3360, 0.0598, 0.0130, 0.0102, 0.0180, 0.0274], + device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0233, 0.0268, 0.0261, 0.0154, 0.0151, 0.0184, 0.0198], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:57:14,571 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=51766.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:57:17,817 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0884, 1.4131, 1.0993, 1.2897, 1.4284, 1.1825, 1.2055, 0.9331], + device='cuda:1'), covar=tensor([0.0127, 0.0105, 0.0289, 0.0123, 0.0101, 0.0101, 0.0253, 0.0187], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0024, 0.0023, 0.0024, 0.0023, 0.0023, 0.0025, 0.0031], + device='cuda:1'), out_proj_covar=tensor([2.9013e-05, 2.7388e-05, 2.7037e-05, 2.6718e-05, 2.6823e-05, 2.5960e-05, + 2.8406e-05, 3.5259e-05], device='cuda:1') +2023-03-21 00:57:23,824 INFO [train.py:901] (1/2) Epoch 19, batch 950, loss[loss=0.1677, simple_loss=0.2517, pruned_loss=0.04182, over 7260.00 frames. ], tot_loss[loss=0.1552, simple_loss=0.2318, pruned_loss=0.03929, over 1428525.17 frames. ], batch size: 64, lr: 8.72e-03, grad_scale: 8.0 +2023-03-21 00:57:23,974 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1246, 2.5596, 2.1243, 3.8235, 1.5916, 3.4772, 1.4230, 3.2383], + device='cuda:1'), covar=tensor([0.0118, 0.0930, 0.1806, 0.0105, 0.3888, 0.0178, 0.1267, 0.0331], + device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0262, 0.0287, 0.0174, 0.0274, 0.0191, 0.0260, 0.0223], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 00:57:24,318 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 00:57:37,516 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51808.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:57:44,393 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.232e+02 2.043e+02 2.485e+02 2.996e+02 5.664e+02, threshold=4.969e+02, percent-clipped=1.0 +2023-03-21 00:57:47,468 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51828.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 00:57:48,378 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 00:57:49,834 INFO [train.py:901] (1/2) Epoch 19, batch 1000, loss[loss=0.1358, simple_loss=0.2052, pruned_loss=0.03319, over 7191.00 frames. ], tot_loss[loss=0.1549, simple_loss=0.2312, pruned_loss=0.03929, over 1430635.16 frames. ], batch size: 39, lr: 8.71e-03, grad_scale: 8.0 +2023-03-21 00:57:56,992 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2390, 4.3950, 4.1506, 4.2862, 4.1290, 4.4335, 4.6499, 4.7021], + device='cuda:1'), covar=tensor([0.0181, 0.0134, 0.0175, 0.0167, 0.0254, 0.0195, 0.0198, 0.0171], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0105, 0.0099, 0.0106, 0.0100, 0.0090, 0.0087, 0.0082], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:57:59,012 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51851.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:58:10,169 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 00:58:12,179 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=51876.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 00:58:16,245 INFO [train.py:901] (1/2) Epoch 19, batch 1050, loss[loss=0.1614, simple_loss=0.2363, pruned_loss=0.04325, over 7337.00 frames. ], tot_loss[loss=0.1547, simple_loss=0.2309, pruned_loss=0.03927, over 1432749.25 frames. ], batch size: 63, lr: 8.71e-03, grad_scale: 8.0 +2023-03-21 00:58:31,811 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 00:58:35,842 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.265e+02 2.062e+02 2.448e+02 3.155e+02 8.388e+02, threshold=4.897e+02, percent-clipped=3.0 +2023-03-21 00:58:35,877 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 00:58:36,527 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9847, 2.6776, 3.0416, 2.8723, 2.9590, 2.7744, 2.3988, 2.8242], + device='cuda:1'), covar=tensor([0.1362, 0.0885, 0.0968, 0.1793, 0.1058, 0.1042, 0.3309, 0.1820], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0052, 0.0039, 0.0040, 0.0039, 0.0037, 0.0055, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 00:58:41,365 INFO [train.py:901] (1/2) Epoch 19, batch 1100, loss[loss=0.1359, simple_loss=0.214, pruned_loss=0.02889, over 7321.00 frames. ], tot_loss[loss=0.1551, simple_loss=0.2314, pruned_loss=0.03935, over 1435405.04 frames. ], batch size: 75, lr: 8.70e-03, grad_scale: 8.0 +2023-03-21 00:58:50,160 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.0641, 0.8254, 1.2590, 1.4872, 1.3631, 1.6593, 1.0684, 1.5108], + device='cuda:1'), covar=tensor([0.2810, 0.3392, 0.1797, 0.0908, 0.1567, 0.1091, 0.1954, 0.1517], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0059, 0.0045, 0.0042, 0.0045, 0.0044, 0.0065, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 00:58:54,639 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4200, 3.6352, 3.3154, 3.5655, 3.2294, 3.5474, 3.8559, 3.8601], + device='cuda:1'), covar=tensor([0.0258, 0.0181, 0.0262, 0.0212, 0.0358, 0.0410, 0.0287, 0.0231], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0106, 0.0100, 0.0106, 0.0102, 0.0090, 0.0088, 0.0084], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:59:05,191 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 00:59:05,689 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 00:59:07,711 INFO [train.py:901] (1/2) Epoch 19, batch 1150, loss[loss=0.1519, simple_loss=0.2269, pruned_loss=0.03848, over 7308.00 frames. ], tot_loss[loss=0.1547, simple_loss=0.2311, pruned_loss=0.03909, over 1433989.18 frames. ], batch size: 83, lr: 8.70e-03, grad_scale: 8.0 +2023-03-21 00:59:21,586 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 00:59:22,066 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 00:59:26,690 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2989, 4.5476, 4.2219, 4.4449, 4.0692, 4.3858, 4.7283, 4.8108], + device='cuda:1'), covar=tensor([0.0174, 0.0113, 0.0163, 0.0150, 0.0325, 0.0244, 0.0186, 0.0135], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0104, 0.0099, 0.0106, 0.0101, 0.0090, 0.0087, 0.0082], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 00:59:31,081 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.626e+02 2.216e+02 2.556e+02 3.107e+02 7.950e+02, threshold=5.111e+02, percent-clipped=3.0 +2023-03-21 00:59:37,237 INFO [train.py:901] (1/2) Epoch 19, batch 1200, loss[loss=0.1465, simple_loss=0.2369, pruned_loss=0.02804, over 7279.00 frames. ], tot_loss[loss=0.1554, simple_loss=0.2321, pruned_loss=0.03937, over 1437946.38 frames. ], batch size: 77, lr: 8.70e-03, grad_scale: 8.0 +2023-03-21 00:59:39,909 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52038.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:59:40,911 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52040.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:59:44,735 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52048.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 00:59:46,566 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 +2023-03-21 00:59:54,778 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 01:00:02,803 INFO [train.py:901] (1/2) Epoch 19, batch 1250, loss[loss=0.1322, simple_loss=0.2087, pruned_loss=0.02784, over 7143.00 frames. ], tot_loss[loss=0.1558, simple_loss=0.2324, pruned_loss=0.03963, over 1440598.87 frames. ], batch size: 39, lr: 8.69e-03, grad_scale: 8.0 +2023-03-21 01:00:04,380 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=52086.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:00:07,932 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3137, 1.5364, 1.3900, 1.4904, 1.7536, 1.4675, 1.5091, 1.3158], + device='cuda:1'), covar=tensor([0.0152, 0.0155, 0.0215, 0.0148, 0.0061, 0.0106, 0.0239, 0.0160], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0024, 0.0024, 0.0024, 0.0023, 0.0023, 0.0025, 0.0031], + device='cuda:1'), out_proj_covar=tensor([2.9237e-05, 2.7338e-05, 2.7551e-05, 2.7245e-05, 2.6627e-05, 2.6041e-05, + 2.8249e-05, 3.6069e-05], device='cuda:1') +2023-03-21 01:00:09,380 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=52096.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:00:12,015 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52101.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:00:12,916 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52103.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:00:18,387 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 01:00:22,540 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4955, 3.7384, 3.5254, 3.6715, 3.3823, 3.4992, 3.9682, 3.9786], + device='cuda:1'), covar=tensor([0.0230, 0.0150, 0.0232, 0.0184, 0.0383, 0.0412, 0.0259, 0.0197], + device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0103, 0.0098, 0.0104, 0.0099, 0.0089, 0.0086, 0.0081], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:00:22,905 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.050e+02 2.393e+02 2.856e+02 5.508e+02, threshold=4.786e+02, percent-clipped=1.0 +2023-03-21 01:00:22,936 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 01:00:24,403 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 01:00:28,082 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5250, 3.5138, 2.4079, 4.0314, 3.1015, 3.5800, 2.0282, 2.2858], + device='cuda:1'), covar=tensor([0.0310, 0.0557, 0.1981, 0.0404, 0.0334, 0.0534, 0.2633, 0.1664], + device='cuda:1'), in_proj_covar=tensor([0.0261, 0.0241, 0.0300, 0.0255, 0.0264, 0.0255, 0.0262, 0.0275], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 01:00:28,410 INFO [train.py:901] (1/2) Epoch 19, batch 1300, loss[loss=0.1524, simple_loss=0.229, pruned_loss=0.0379, over 7233.00 frames. ], tot_loss[loss=0.1558, simple_loss=0.232, pruned_loss=0.03975, over 1440503.86 frames. ], batch size: 45, lr: 8.69e-03, grad_scale: 8.0 +2023-03-21 01:00:38,309 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52151.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:00:47,148 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 01:00:50,078 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 01:00:53,670 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 01:00:54,142 INFO [train.py:901] (1/2) Epoch 19, batch 1350, loss[loss=0.155, simple_loss=0.2337, pruned_loss=0.03815, over 7274.00 frames. ], tot_loss[loss=0.1552, simple_loss=0.2315, pruned_loss=0.03948, over 1441547.74 frames. ], batch size: 52, lr: 8.68e-03, grad_scale: 8.0 +2023-03-21 01:00:59,257 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52193.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:01:02,298 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=52199.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:01:04,546 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 01:01:14,634 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.452e+02 2.232e+02 2.578e+02 3.253e+02 9.285e+02, threshold=5.156e+02, percent-clipped=5.0 +2023-03-21 01:01:16,838 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9907, 2.4033, 1.8097, 2.9216, 2.7733, 2.8526, 2.5455, 2.3969], + device='cuda:1'), covar=tensor([0.1795, 0.0827, 0.3070, 0.0536, 0.0117, 0.0085, 0.0182, 0.0259], + device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0231, 0.0264, 0.0261, 0.0154, 0.0151, 0.0185, 0.0197], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:01:20,779 INFO [train.py:901] (1/2) Epoch 19, batch 1400, loss[loss=0.165, simple_loss=0.2437, pruned_loss=0.04316, over 7251.00 frames. ], tot_loss[loss=0.1549, simple_loss=0.2312, pruned_loss=0.03933, over 1440448.19 frames. ], batch size: 64, lr: 8.68e-03, grad_scale: 8.0 +2023-03-21 01:01:31,282 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52254.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:01:37,690 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 01:01:45,617 INFO [train.py:901] (1/2) Epoch 19, batch 1450, loss[loss=0.1843, simple_loss=0.2646, pruned_loss=0.05201, over 7237.00 frames. ], tot_loss[loss=0.155, simple_loss=0.2312, pruned_loss=0.03937, over 1441025.07 frames. ], batch size: 55, lr: 8.68e-03, grad_scale: 8.0 +2023-03-21 01:01:53,852 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8459, 3.1420, 2.5244, 2.9748, 2.8248, 2.4384, 3.0162, 3.0632], + device='cuda:1'), covar=tensor([0.1378, 0.0941, 0.1459, 0.1356, 0.1159, 0.1825, 0.1004, 0.0845], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0044, 0.0052, 0.0046, 0.0044, 0.0047, 0.0047, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:02:00,957 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52312.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:02:01,377 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 01:02:06,382 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.471e+02 2.084e+02 2.470e+02 2.878e+02 4.955e+02, threshold=4.940e+02, percent-clipped=0.0 +2023-03-21 01:02:09,123 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1704, 2.6145, 2.3729, 3.6952, 1.6419, 3.5087, 1.3593, 3.1860], + device='cuda:1'), covar=tensor([0.0106, 0.0887, 0.1555, 0.0101, 0.3769, 0.0106, 0.1034, 0.0291], + device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0260, 0.0285, 0.0176, 0.0269, 0.0187, 0.0255, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:02:12,004 INFO [train.py:901] (1/2) Epoch 19, batch 1500, loss[loss=0.1409, simple_loss=0.2225, pruned_loss=0.02965, over 7374.00 frames. ], tot_loss[loss=0.1546, simple_loss=0.231, pruned_loss=0.0391, over 1441836.16 frames. ], batch size: 63, lr: 8.67e-03, grad_scale: 8.0 +2023-03-21 01:02:17,970 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 01:02:25,673 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.9073, 4.3611, 4.3874, 4.8339, 4.8135, 4.8217, 4.2375, 4.4087], + device='cuda:1'), covar=tensor([0.0727, 0.2425, 0.2151, 0.1071, 0.0727, 0.1267, 0.0791, 0.0938], + device='cuda:1'), in_proj_covar=tensor([0.0213, 0.0324, 0.0259, 0.0261, 0.0190, 0.0319, 0.0188, 0.0233], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:02:32,298 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52373.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:02:37,809 INFO [train.py:901] (1/2) Epoch 19, batch 1550, loss[loss=0.1805, simple_loss=0.2495, pruned_loss=0.05574, over 7319.00 frames. ], tot_loss[loss=0.154, simple_loss=0.2305, pruned_loss=0.03878, over 1441161.05 frames. ], batch size: 49, lr: 8.67e-03, grad_scale: 16.0 +2023-03-21 01:02:42,975 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6013, 2.8978, 2.5020, 2.7918, 2.6434, 2.5307, 2.8564, 2.9230], + device='cuda:1'), covar=tensor([0.1016, 0.0848, 0.1226, 0.1376, 0.1403, 0.0644, 0.1021, 0.0666], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0045, 0.0052, 0.0046, 0.0045, 0.0047, 0.0047, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:02:43,364 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 01:02:44,426 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52396.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:02:44,498 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2348, 2.5400, 2.2544, 2.4811, 2.4134, 2.3322, 2.5248, 2.4471], + device='cuda:1'), covar=tensor([0.0866, 0.0605, 0.1097, 0.0630, 0.0857, 0.0436, 0.0769, 0.0971], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0045, 0.0052, 0.0046, 0.0045, 0.0047, 0.0047, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:02:48,262 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52403.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:02:58,271 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.373e+02 1.980e+02 2.348e+02 2.883e+02 5.696e+02, threshold=4.696e+02, percent-clipped=3.0 +2023-03-21 01:02:58,944 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8961, 2.5185, 2.4420, 2.3605, 2.5641, 2.2945, 2.2856, 1.7412], + device='cuda:1'), covar=tensor([0.0445, 0.0267, 0.0109, 0.0148, 0.0362, 0.0219, 0.0188, 0.0281], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0027, 0.0024, 0.0024, 0.0025, 0.0024, 0.0028, 0.0027], + device='cuda:1'), out_proj_covar=tensor([6.8793e-05, 6.7961e-05, 6.0593e-05, 5.9375e-05, 6.4538e-05, 6.1507e-05, + 6.8050e-05, 6.9241e-05], device='cuda:1') +2023-03-21 01:03:03,900 INFO [train.py:901] (1/2) Epoch 19, batch 1600, loss[loss=0.1442, simple_loss=0.2303, pruned_loss=0.02902, over 7307.00 frames. ], tot_loss[loss=0.1545, simple_loss=0.231, pruned_loss=0.03894, over 1441139.20 frames. ], batch size: 80, lr: 8.66e-03, grad_scale: 16.0 +2023-03-21 01:03:10,040 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2906, 1.4484, 1.3057, 1.4934, 1.5476, 1.3497, 1.4370, 1.0641], + device='cuda:1'), covar=tensor([0.0130, 0.0129, 0.0296, 0.0097, 0.0111, 0.0103, 0.0089, 0.0142], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0024, 0.0024, 0.0025, 0.0024, 0.0024, 0.0025, 0.0032], + device='cuda:1'), out_proj_covar=tensor([2.9530e-05, 2.7837e-05, 2.7724e-05, 2.7447e-05, 2.8056e-05, 2.6649e-05, + 2.8882e-05, 3.7325e-05], device='cuda:1') +2023-03-21 01:03:10,516 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8516, 2.3286, 2.2956, 1.9744, 2.3878, 2.2088, 2.1231, 1.5411], + device='cuda:1'), covar=tensor([0.0361, 0.0232, 0.0085, 0.0148, 0.0289, 0.0230, 0.0203, 0.0245], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0026, 0.0024, 0.0024, 0.0025, 0.0024, 0.0028, 0.0027], + device='cuda:1'), out_proj_covar=tensor([6.8789e-05, 6.7463e-05, 6.0558e-05, 5.9367e-05, 6.4323e-05, 6.1314e-05, + 6.7593e-05, 6.9120e-05], device='cuda:1') +2023-03-21 01:03:12,990 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=52451.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:03:13,928 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 01:03:14,962 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 01:03:17,964 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 01:03:27,015 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 01:03:29,517 INFO [train.py:901] (1/2) Epoch 19, batch 1650, loss[loss=0.1499, simple_loss=0.2315, pruned_loss=0.03419, over 7241.00 frames. ], tot_loss[loss=0.1553, simple_loss=0.232, pruned_loss=0.03928, over 1439756.79 frames. ], batch size: 89, lr: 8.66e-03, grad_scale: 16.0 +2023-03-21 01:03:31,490 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 01:03:39,815 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 01:03:49,858 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.218e+02 2.238e+02 2.612e+02 3.033e+02 5.856e+02, threshold=5.224e+02, percent-clipped=2.0 +2023-03-21 01:03:55,454 INFO [train.py:901] (1/2) Epoch 19, batch 1700, loss[loss=0.1496, simple_loss=0.2302, pruned_loss=0.03451, over 7253.00 frames. ], tot_loss[loss=0.1543, simple_loss=0.231, pruned_loss=0.03877, over 1439730.67 frames. ], batch size: 47, lr: 8.65e-03, grad_scale: 8.0 +2023-03-21 01:03:55,468 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 01:03:59,512 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 01:04:04,311 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52549.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:04:07,383 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52555.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 01:04:10,825 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 01:04:21,640 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8786, 2.4161, 1.8635, 3.0909, 2.3968, 2.9492, 2.1087, 2.5753], + device='cuda:1'), covar=tensor([0.1861, 0.0866, 0.3215, 0.0512, 0.0101, 0.0101, 0.0156, 0.0312], + device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0236, 0.0271, 0.0264, 0.0156, 0.0155, 0.0186, 0.0201], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:04:21,986 INFO [train.py:901] (1/2) Epoch 19, batch 1750, loss[loss=0.1509, simple_loss=0.2293, pruned_loss=0.03625, over 7273.00 frames. ], tot_loss[loss=0.1545, simple_loss=0.2314, pruned_loss=0.0388, over 1441032.38 frames. ], batch size: 52, lr: 8.65e-03, grad_scale: 8.0 +2023-03-21 01:04:34,001 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 01:04:34,996 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 01:04:39,159 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52616.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 01:04:42,429 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.001e+02 2.278e+02 2.833e+02 4.492e+02, threshold=4.555e+02, percent-clipped=0.0 +2023-03-21 01:04:47,522 INFO [train.py:901] (1/2) Epoch 19, batch 1800, loss[loss=0.1527, simple_loss=0.2248, pruned_loss=0.04034, over 7269.00 frames. ], tot_loss[loss=0.1534, simple_loss=0.2304, pruned_loss=0.03824, over 1439678.07 frames. ], batch size: 52, lr: 8.65e-03, grad_scale: 8.0 +2023-03-21 01:04:57,743 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 01:05:06,610 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52668.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:05:11,501 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 01:05:13,957 INFO [train.py:901] (1/2) Epoch 19, batch 1850, loss[loss=0.1554, simple_loss=0.2425, pruned_loss=0.03418, over 7258.00 frames. ], tot_loss[loss=0.1528, simple_loss=0.2296, pruned_loss=0.038, over 1437872.88 frames. ], batch size: 64, lr: 8.64e-03, grad_scale: 8.0 +2023-03-21 01:05:20,365 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 01:05:20,460 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52696.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:05:34,425 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.453e+02 2.173e+02 2.520e+02 2.887e+02 6.459e+02, threshold=5.039e+02, percent-clipped=3.0 +2023-03-21 01:05:37,976 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 01:05:39,491 INFO [train.py:901] (1/2) Epoch 19, batch 1900, loss[loss=0.1214, simple_loss=0.1982, pruned_loss=0.02234, over 7188.00 frames. ], tot_loss[loss=0.153, simple_loss=0.2299, pruned_loss=0.03801, over 1441273.26 frames. ], batch size: 39, lr: 8.64e-03, grad_scale: 8.0 +2023-03-21 01:05:45,152 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=52744.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:05:47,896 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3478, 3.0394, 3.1875, 3.0568, 2.6947, 2.5320, 3.4200, 2.5157], + device='cuda:1'), covar=tensor([0.0333, 0.0318, 0.0355, 0.0317, 0.0441, 0.0616, 0.0399, 0.1297], + device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0326, 0.0263, 0.0346, 0.0310, 0.0302, 0.0334, 0.0296], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:05:49,845 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52752.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:05:57,527 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5503, 3.1686, 3.1098, 3.2213, 2.8372, 2.6607, 3.6616, 2.5045], + device='cuda:1'), covar=tensor([0.0390, 0.0362, 0.0498, 0.0382, 0.0566, 0.0743, 0.0480, 0.1887], + device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0329, 0.0265, 0.0350, 0.0314, 0.0304, 0.0337, 0.0299], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:06:02,377 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 01:06:05,397 INFO [train.py:901] (1/2) Epoch 19, batch 1950, loss[loss=0.1414, simple_loss=0.2185, pruned_loss=0.03214, over 7356.00 frames. ], tot_loss[loss=0.1535, simple_loss=0.2304, pruned_loss=0.03834, over 1439319.86 frames. ], batch size: 63, lr: 8.63e-03, grad_scale: 8.0 +2023-03-21 01:06:13,035 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 01:06:18,145 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 01:06:18,650 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 01:06:21,293 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52813.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:06:26,200 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 2.025e+02 2.271e+02 2.764e+02 6.029e+02, threshold=4.543e+02, percent-clipped=1.0 +2023-03-21 01:06:28,625 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 01:06:31,312 INFO [train.py:901] (1/2) Epoch 19, batch 2000, loss[loss=0.1773, simple_loss=0.2478, pruned_loss=0.0534, over 7269.00 frames. ], tot_loss[loss=0.1531, simple_loss=0.2299, pruned_loss=0.03813, over 1436177.17 frames. ], batch size: 77, lr: 8.63e-03, grad_scale: 8.0 +2023-03-21 01:06:35,576 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 01:06:40,205 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52849.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:06:45,295 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1394, 1.4012, 1.2571, 1.3038, 1.2197, 1.1452, 1.1490, 0.9546], + device='cuda:1'), covar=tensor([0.0129, 0.0074, 0.0135, 0.0082, 0.0060, 0.0114, 0.0114, 0.0117], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0024, 0.0024, 0.0024, 0.0023, 0.0023, 0.0025, 0.0032], + device='cuda:1'), out_proj_covar=tensor([2.9402e-05, 2.7324e-05, 2.7641e-05, 2.7053e-05, 2.7339e-05, 2.6133e-05, + 2.8510e-05, 3.7138e-05], device='cuda:1') +2023-03-21 01:06:46,633 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 01:06:55,199 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 01:06:57,195 INFO [train.py:901] (1/2) Epoch 19, batch 2050, loss[loss=0.1434, simple_loss=0.2253, pruned_loss=0.03075, over 7285.00 frames. ], tot_loss[loss=0.1521, simple_loss=0.2287, pruned_loss=0.03773, over 1436363.52 frames. ], batch size: 77, lr: 8.63e-03, grad_scale: 8.0 +2023-03-21 01:07:04,810 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=52897.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:07:11,921 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52911.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 01:07:18,354 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.311e+02 2.118e+02 2.490e+02 3.094e+02 4.672e+02, threshold=4.979e+02, percent-clipped=1.0 +2023-03-21 01:07:23,058 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6230, 3.2502, 3.1487, 3.2728, 2.8208, 2.6934, 3.5845, 2.6479], + device='cuda:1'), covar=tensor([0.0362, 0.0293, 0.0492, 0.0344, 0.0701, 0.0874, 0.0379, 0.1755], + device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0328, 0.0265, 0.0347, 0.0311, 0.0303, 0.0332, 0.0297], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:07:23,358 INFO [train.py:901] (1/2) Epoch 19, batch 2100, loss[loss=0.1466, simple_loss=0.2241, pruned_loss=0.03454, over 7339.00 frames. ], tot_loss[loss=0.1527, simple_loss=0.2295, pruned_loss=0.03797, over 1435596.77 frames. ], batch size: 51, lr: 8.62e-03, grad_scale: 8.0 +2023-03-21 01:07:28,819 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 01:07:31,829 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 01:07:40,787 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52968.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:07:48,867 INFO [train.py:901] (1/2) Epoch 19, batch 2150, loss[loss=0.1618, simple_loss=0.2452, pruned_loss=0.03924, over 6725.00 frames. ], tot_loss[loss=0.1536, simple_loss=0.2305, pruned_loss=0.03837, over 1436949.86 frames. ], batch size: 107, lr: 8.62e-03, grad_scale: 8.0 +2023-03-21 01:08:06,593 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=53016.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:08:09,986 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 9.314e+01 1.872e+02 2.271e+02 2.909e+02 5.856e+02, threshold=4.542e+02, percent-clipped=2.0 +2023-03-21 01:08:15,049 INFO [train.py:901] (1/2) Epoch 19, batch 2200, loss[loss=0.1536, simple_loss=0.2314, pruned_loss=0.03793, over 7278.00 frames. ], tot_loss[loss=0.1539, simple_loss=0.2309, pruned_loss=0.0385, over 1435646.28 frames. ], batch size: 68, lr: 8.61e-03, grad_scale: 8.0 +2023-03-21 01:08:18,088 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 01:08:18,233 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1793, 1.4408, 1.2666, 1.3577, 1.3356, 1.2681, 1.3145, 1.1283], + device='cuda:1'), covar=tensor([0.0147, 0.0087, 0.0182, 0.0102, 0.0073, 0.0088, 0.0091, 0.0105], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0024, 0.0024, 0.0024, 0.0023, 0.0023, 0.0025, 0.0032], + device='cuda:1'), out_proj_covar=tensor([2.9174e-05, 2.7315e-05, 2.7661e-05, 2.6879e-05, 2.7363e-05, 2.6100e-05, + 2.8663e-05, 3.6720e-05], device='cuda:1') +2023-03-21 01:08:27,370 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53056.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:08:40,776 INFO [train.py:901] (1/2) Epoch 19, batch 2250, loss[loss=0.1626, simple_loss=0.2365, pruned_loss=0.04433, over 7304.00 frames. ], tot_loss[loss=0.154, simple_loss=0.2311, pruned_loss=0.03848, over 1439907.21 frames. ], batch size: 80, lr: 8.61e-03, grad_scale: 8.0 +2023-03-21 01:08:41,137 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.42 vs. limit=5.0 +2023-03-21 01:08:50,197 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1005, 2.6773, 3.1710, 2.7315, 3.2644, 2.6762, 2.5030, 3.0696], + device='cuda:1'), covar=tensor([0.1289, 0.0791, 0.1338, 0.2550, 0.0825, 0.1268, 0.3254, 0.1356], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0053, 0.0040, 0.0041, 0.0041, 0.0038, 0.0057, 0.0043], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:08:52,106 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 01:08:52,590 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 01:08:54,153 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53108.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:08:54,184 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0853, 4.2437, 4.0087, 4.3121, 3.7754, 4.3155, 4.5471, 4.5642], + device='cuda:1'), covar=tensor([0.0205, 0.0143, 0.0178, 0.0152, 0.0362, 0.0189, 0.0220, 0.0153], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0105, 0.0101, 0.0107, 0.0100, 0.0088, 0.0086, 0.0080], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:08:58,796 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53117.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:08:59,058 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 +2023-03-21 01:09:01,688 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.241e+02 1.920e+02 2.278e+02 2.763e+02 7.620e+02, threshold=4.557e+02, percent-clipped=1.0 +2023-03-21 01:09:05,263 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 01:09:06,735 INFO [train.py:901] (1/2) Epoch 19, batch 2300, loss[loss=0.1702, simple_loss=0.253, pruned_loss=0.04367, over 7207.00 frames. ], tot_loss[loss=0.1534, simple_loss=0.2302, pruned_loss=0.03835, over 1437268.46 frames. ], batch size: 93, lr: 8.61e-03, grad_scale: 8.0 +2023-03-21 01:09:32,434 INFO [train.py:901] (1/2) Epoch 19, batch 2350, loss[loss=0.1672, simple_loss=0.2468, pruned_loss=0.04378, over 7304.00 frames. ], tot_loss[loss=0.1539, simple_loss=0.2306, pruned_loss=0.0386, over 1438118.54 frames. ], batch size: 86, lr: 8.60e-03, grad_scale: 8.0 +2023-03-21 01:09:47,694 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53211.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 01:09:52,487 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 01:09:53,469 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.463e+02 2.164e+02 2.497e+02 3.144e+02 6.644e+02, threshold=4.994e+02, percent-clipped=6.0 +2023-03-21 01:09:59,150 INFO [train.py:901] (1/2) Epoch 19, batch 2400, loss[loss=0.144, simple_loss=0.2288, pruned_loss=0.02961, over 7227.00 frames. ], tot_loss[loss=0.154, simple_loss=0.2309, pruned_loss=0.03853, over 1439331.65 frames. ], batch size: 93, lr: 8.60e-03, grad_scale: 8.0 +2023-03-21 01:09:59,673 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 01:10:03,342 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53241.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:10:10,347 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 01:10:12,321 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=53259.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 01:10:13,240 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 01:10:25,029 INFO [train.py:901] (1/2) Epoch 19, batch 2450, loss[loss=0.1181, simple_loss=0.1787, pruned_loss=0.02874, over 6087.00 frames. ], tot_loss[loss=0.1534, simple_loss=0.2306, pruned_loss=0.03814, over 1439080.44 frames. ], batch size: 26, lr: 8.59e-03, grad_scale: 8.0 +2023-03-21 01:10:31,939 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-03-21 01:10:34,770 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53302.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:10:35,787 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53304.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:10:40,238 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 01:10:41,401 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53315.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:10:45,883 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.369e+02 1.898e+02 2.232e+02 2.806e+02 4.781e+02, threshold=4.465e+02, percent-clipped=0.0 +2023-03-21 01:10:47,016 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8188, 3.0468, 2.4187, 2.8077, 2.8330, 2.5366, 2.8372, 2.7740], + device='cuda:1'), covar=tensor([0.0661, 0.0847, 0.0976, 0.1164, 0.1350, 0.0589, 0.1335, 0.1009], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0045, 0.0052, 0.0047, 0.0046, 0.0047, 0.0047, 0.0043], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:10:50,953 INFO [train.py:901] (1/2) Epoch 19, batch 2500, loss[loss=0.1271, simple_loss=0.1964, pruned_loss=0.02886, over 6936.00 frames. ], tot_loss[loss=0.1535, simple_loss=0.2309, pruned_loss=0.03808, over 1442317.31 frames. ], batch size: 35, lr: 8.59e-03, grad_scale: 8.0 +2023-03-21 01:11:05,976 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 01:11:07,711 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53365.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:11:08,757 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53367.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:11:13,291 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53376.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:11:16,615 INFO [train.py:901] (1/2) Epoch 19, batch 2550, loss[loss=0.1652, simple_loss=0.2418, pruned_loss=0.0443, over 7270.00 frames. ], tot_loss[loss=0.154, simple_loss=0.2313, pruned_loss=0.03829, over 1445143.01 frames. ], batch size: 64, lr: 8.59e-03, grad_scale: 8.0 +2023-03-21 01:11:29,599 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53408.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:11:32,137 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53412.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:11:37,568 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.453e+02 2.065e+02 2.474e+02 2.931e+02 4.379e+02, threshold=4.948e+02, percent-clipped=0.0 +2023-03-21 01:11:40,250 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53428.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 01:11:42,558 INFO [train.py:901] (1/2) Epoch 19, batch 2600, loss[loss=0.1892, simple_loss=0.2642, pruned_loss=0.05706, over 6641.00 frames. ], tot_loss[loss=0.1539, simple_loss=0.2311, pruned_loss=0.03837, over 1445428.37 frames. ], batch size: 106, lr: 8.58e-03, grad_scale: 8.0 +2023-03-21 01:11:46,869 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-03-21 01:11:48,084 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6471, 1.4837, 1.6747, 2.1095, 1.8191, 2.1287, 1.8127, 1.9635], + device='cuda:1'), covar=tensor([0.2225, 0.2978, 0.2775, 0.0949, 0.2606, 0.4613, 0.2579, 0.2140], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0059, 0.0046, 0.0042, 0.0046, 0.0046, 0.0066, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 01:11:51,564 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3668, 1.1437, 1.5866, 1.8245, 1.7929, 1.9040, 1.5199, 1.7311], + device='cuda:1'), covar=tensor([0.1486, 0.2129, 0.2432, 0.0725, 0.1829, 0.2606, 0.1553, 0.2244], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0059, 0.0046, 0.0042, 0.0046, 0.0046, 0.0066, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 01:11:53,993 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=53456.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:12:07,311 INFO [train.py:901] (1/2) Epoch 19, batch 2650, loss[loss=0.1535, simple_loss=0.2359, pruned_loss=0.03554, over 7326.00 frames. ], tot_loss[loss=0.154, simple_loss=0.2311, pruned_loss=0.03841, over 1444901.96 frames. ], batch size: 83, lr: 8.58e-03, grad_scale: 8.0 +2023-03-21 01:12:27,467 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.393e+02 1.972e+02 2.315e+02 2.811e+02 6.076e+02, threshold=4.629e+02, percent-clipped=2.0 +2023-03-21 01:12:32,414 INFO [train.py:901] (1/2) Epoch 19, batch 2700, loss[loss=0.1589, simple_loss=0.2374, pruned_loss=0.04021, over 7254.00 frames. ], tot_loss[loss=0.1541, simple_loss=0.2313, pruned_loss=0.03845, over 1442755.93 frames. ], batch size: 89, lr: 8.57e-03, grad_scale: 8.0 +2023-03-21 01:12:33,019 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2052, 4.0573, 3.6161, 3.4796, 3.1600, 2.5169, 1.7780, 4.2028], + device='cuda:1'), covar=tensor([0.0033, 0.0041, 0.0086, 0.0068, 0.0122, 0.0430, 0.0555, 0.0047], + device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0074, 0.0093, 0.0080, 0.0103, 0.0118, 0.0118, 0.0088], + device='cuda:1'), out_proj_covar=tensor([1.0589e-04, 9.9169e-05, 1.1891e-04, 1.0551e-04, 1.2949e-04, 1.5016e-04, + 1.5116e-04, 1.0903e-04], device='cuda:1') +2023-03-21 01:12:45,984 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 +2023-03-21 01:12:47,145 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.1933, 4.7571, 4.7380, 5.1930, 5.1544, 5.1470, 4.4540, 4.7584], + device='cuda:1'), covar=tensor([0.0727, 0.2384, 0.2118, 0.0940, 0.0796, 0.1222, 0.0730, 0.0967], + device='cuda:1'), in_proj_covar=tensor([0.0214, 0.0328, 0.0265, 0.0262, 0.0192, 0.0322, 0.0191, 0.0234], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:12:49,300 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-21 01:12:52,512 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4719, 3.6461, 3.4615, 3.7022, 3.2007, 3.5402, 3.8302, 3.8708], + device='cuda:1'), covar=tensor([0.0285, 0.0214, 0.0238, 0.0207, 0.0521, 0.0374, 0.0309, 0.0240], + device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0108, 0.0103, 0.0109, 0.0104, 0.0090, 0.0088, 0.0084], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:12:56,803 INFO [train.py:901] (1/2) Epoch 19, batch 2750, loss[loss=0.1422, simple_loss=0.218, pruned_loss=0.0332, over 7287.00 frames. ], tot_loss[loss=0.1536, simple_loss=0.2307, pruned_loss=0.03824, over 1442990.51 frames. ], batch size: 66, lr: 8.57e-03, grad_scale: 8.0 +2023-03-21 01:13:03,829 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53597.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:13:16,768 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 2.108e+02 2.376e+02 2.842e+02 5.465e+02, threshold=4.753e+02, percent-clipped=1.0 +2023-03-21 01:13:20,904 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 +2023-03-21 01:13:21,587 INFO [train.py:901] (1/2) Epoch 19, batch 2800, loss[loss=0.1586, simple_loss=0.2363, pruned_loss=0.04048, over 7275.00 frames. ], tot_loss[loss=0.1536, simple_loss=0.2305, pruned_loss=0.03831, over 1443088.86 frames. ], batch size: 70, lr: 8.57e-03, grad_scale: 8.0 +2023-03-21 01:13:47,337 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 01:13:47,389 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 01:13:47,412 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 01:13:47,430 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 01:13:47,602 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 01:13:47,616 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 01:13:47,628 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 01:13:47,728 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 01:13:47,802 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 01:13:47,816 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 01:13:47,833 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 01:13:48,341 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 01:13:48,346 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 01:13:48,355 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 01:13:48,495 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 01:13:48,498 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 01:13:48,555 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 01:13:55,244 INFO [train.py:901] (1/2) Epoch 20, batch 0, loss[loss=0.1603, simple_loss=0.2414, pruned_loss=0.03958, over 7126.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2414, pruned_loss=0.03958, over 7126.00 frames. ], batch size: 98, lr: 8.36e-03, grad_scale: 8.0 +2023-03-21 01:13:55,245 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 01:14:08,119 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5194, 4.1040, 3.8825, 4.0377, 4.1585, 4.0045, 4.1001, 3.7784], + device='cuda:1'), covar=tensor([0.0103, 0.0123, 0.0167, 0.0170, 0.0379, 0.0117, 0.0198, 0.0211], + device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0082, 0.0082, 0.0072, 0.0147, 0.0092, 0.0087, 0.0091], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:14:10,383 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6170, 2.7767, 2.0960, 3.3436, 2.1973, 2.7497, 1.5540, 2.2681], + device='cuda:1'), covar=tensor([0.0257, 0.0622, 0.2089, 0.0382, 0.0390, 0.0507, 0.3099, 0.1336], + device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0245, 0.0299, 0.0254, 0.0266, 0.0257, 0.0261, 0.0275], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 01:14:14,847 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5442, 4.7926, 4.7823, 4.7754, 4.5979, 4.4332, 4.8640, 4.5419], + device='cuda:1'), covar=tensor([0.0340, 0.0338, 0.0350, 0.0414, 0.0318, 0.0270, 0.0252, 0.0486], + device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0220, 0.0160, 0.0162, 0.0131, 0.0196, 0.0169, 0.0130], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:14:16,513 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1256, 4.1806, 4.0213, 4.2499, 4.1585, 4.1876, 4.3296, 4.3758], + device='cuda:1'), covar=tensor([0.0037, 0.0064, 0.0034, 0.0027, 0.0031, 0.0032, 0.0027, 0.0039], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0055, 0.0047, 0.0046, 0.0046, 0.0049, 0.0047, 0.0059], + device='cuda:1'), out_proj_covar=tensor([8.1165e-05, 1.3087e-04, 1.0709e-04, 9.7233e-05, 9.7141e-05, 1.0362e-04, + 1.0929e-04, 1.2865e-04], device='cuda:1') +2023-03-21 01:14:20,364 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3595, 1.4352, 1.3494, 1.3918, 1.2975, 1.3732, 1.3452, 1.1961], + device='cuda:1'), covar=tensor([0.0172, 0.0187, 0.0273, 0.0129, 0.0118, 0.0108, 0.0126, 0.0141], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0024, 0.0024, 0.0024, 0.0024, 0.0023, 0.0025, 0.0032], + device='cuda:1'), out_proj_covar=tensor([2.8948e-05, 2.7764e-05, 2.8005e-05, 2.6843e-05, 2.7822e-05, 2.6139e-05, + 2.8681e-05, 3.7216e-05], device='cuda:1') +2023-03-21 01:14:20,799 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6087, 1.9354, 1.9469, 1.9556, 2.3464, 1.7685, 2.0387, 1.3440], + device='cuda:1'), covar=tensor([0.0341, 0.0427, 0.0294, 0.0135, 0.0312, 0.0805, 0.0166, 0.0317], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0027, 0.0025, 0.0024, 0.0026, 0.0025, 0.0028, 0.0028], + device='cuda:1'), out_proj_covar=tensor([7.0782e-05, 6.9267e-05, 6.2981e-05, 6.1127e-05, 6.6113e-05, 6.4225e-05, + 6.9364e-05, 7.0859e-05], device='cuda:1') +2023-03-21 01:14:21,487 INFO [train.py:935] (1/2) Epoch 20, validation: loss=0.166, simple_loss=0.2535, pruned_loss=0.03929, over 1622729.00 frames. +2023-03-21 01:14:21,488 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 01:14:23,082 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53660.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:14:27,906 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 01:14:28,478 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53671.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:14:37,115 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4986, 2.5971, 2.2819, 3.5106, 1.5504, 3.2860, 1.2870, 2.6174], + device='cuda:1'), covar=tensor([0.0110, 0.0863, 0.1557, 0.0116, 0.3829, 0.0144, 0.1360, 0.0253], + device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0261, 0.0287, 0.0174, 0.0271, 0.0190, 0.0258, 0.0222], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:14:38,506 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 01:14:45,589 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 01:14:46,595 INFO [train.py:901] (1/2) Epoch 20, batch 50, loss[loss=0.1626, simple_loss=0.2446, pruned_loss=0.04027, over 7226.00 frames. ], tot_loss[loss=0.1515, simple_loss=0.2285, pruned_loss=0.03726, over 323053.27 frames. ], batch size: 93, lr: 8.35e-03, grad_scale: 8.0 +2023-03-21 01:14:47,647 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 01:14:49,307 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53712.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:14:50,678 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 01:14:55,338 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 2.061e+02 2.323e+02 2.773e+02 4.898e+02, threshold=4.647e+02, percent-clipped=1.0 +2023-03-21 01:14:55,432 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53723.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 01:15:08,209 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 01:15:08,221 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 01:15:13,129 INFO [train.py:901] (1/2) Epoch 20, batch 100, loss[loss=0.1627, simple_loss=0.2366, pruned_loss=0.04442, over 7357.00 frames. ], tot_loss[loss=0.1524, simple_loss=0.2306, pruned_loss=0.03712, over 572196.14 frames. ], batch size: 73, lr: 8.35e-03, grad_scale: 8.0 +2023-03-21 01:15:14,678 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=53760.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:15:17,913 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3968, 2.6202, 2.2550, 3.5328, 1.6162, 3.3884, 1.4193, 2.8835], + device='cuda:1'), covar=tensor([0.0092, 0.0751, 0.1496, 0.0076, 0.3589, 0.0137, 0.0988, 0.0203], + device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0259, 0.0283, 0.0172, 0.0268, 0.0189, 0.0255, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:15:22,426 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9853, 4.3936, 4.3508, 4.3927, 4.3964, 4.1427, 4.5058, 4.4017], + device='cuda:1'), covar=tensor([0.1082, 0.1234, 0.0970, 0.1021, 0.0732, 0.0900, 0.0950, 0.0975], + device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0221, 0.0161, 0.0163, 0.0130, 0.0198, 0.0172, 0.0130], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:15:38,714 INFO [train.py:901] (1/2) Epoch 20, batch 150, loss[loss=0.1491, simple_loss=0.227, pruned_loss=0.03561, over 7308.00 frames. ], tot_loss[loss=0.1522, simple_loss=0.2304, pruned_loss=0.03702, over 766342.66 frames. ], batch size: 80, lr: 8.35e-03, grad_scale: 8.0 +2023-03-21 01:15:47,349 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.409e+02 1.894e+02 2.222e+02 2.716e+02 9.009e+02, threshold=4.445e+02, percent-clipped=3.0 +2023-03-21 01:16:05,180 INFO [train.py:901] (1/2) Epoch 20, batch 200, loss[loss=0.1624, simple_loss=0.2433, pruned_loss=0.04072, over 6679.00 frames. ], tot_loss[loss=0.1522, simple_loss=0.2307, pruned_loss=0.03679, over 917427.88 frames. ], batch size: 106, lr: 8.34e-03, grad_scale: 8.0 +2023-03-21 01:16:08,189 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 01:16:12,372 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 01:16:16,748 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-21 01:16:18,299 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 01:16:26,211 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53897.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:16:31,784 INFO [train.py:901] (1/2) Epoch 20, batch 250, loss[loss=0.1552, simple_loss=0.2333, pruned_loss=0.0385, over 7280.00 frames. ], tot_loss[loss=0.1525, simple_loss=0.231, pruned_loss=0.03696, over 1034252.00 frames. ], batch size: 52, lr: 8.34e-03, grad_scale: 8.0 +2023-03-21 01:16:31,795 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 01:16:39,791 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 2.005e+02 2.500e+02 2.996e+02 9.929e+02, threshold=5.000e+02, percent-clipped=6.0 +2023-03-21 01:16:50,909 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=53945.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:16:52,837 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 01:16:56,812 INFO [train.py:901] (1/2) Epoch 20, batch 300, loss[loss=0.1552, simple_loss=0.2347, pruned_loss=0.03781, over 7302.00 frames. ], tot_loss[loss=0.1518, simple_loss=0.2297, pruned_loss=0.03697, over 1124377.84 frames. ], batch size: 59, lr: 8.33e-03, grad_scale: 8.0 +2023-03-21 01:16:58,454 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53960.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:17:01,462 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 01:17:04,038 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53971.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:17:23,227 INFO [train.py:901] (1/2) Epoch 20, batch 350, loss[loss=0.1252, simple_loss=0.2019, pruned_loss=0.02423, over 6944.00 frames. ], tot_loss[loss=0.1517, simple_loss=0.2292, pruned_loss=0.03705, over 1196373.47 frames. ], batch size: 35, lr: 8.33e-03, grad_scale: 8.0 +2023-03-21 01:17:23,816 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=54008.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:17:29,356 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=54019.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:17:31,242 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.231e+02 1.892e+02 2.279e+02 2.732e+02 6.693e+02, threshold=4.558e+02, percent-clipped=2.0 +2023-03-21 01:17:31,374 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=54023.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 01:17:37,251 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 01:17:42,537 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-03-21 01:17:48,241 INFO [train.py:901] (1/2) Epoch 20, batch 400, loss[loss=0.1211, simple_loss=0.1941, pruned_loss=0.02403, over 7170.00 frames. ], tot_loss[loss=0.1516, simple_loss=0.2291, pruned_loss=0.03709, over 1250748.89 frames. ], batch size: 39, lr: 8.33e-03, grad_scale: 8.0 +2023-03-21 01:17:55,905 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=54071.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:18:14,575 INFO [train.py:901] (1/2) Epoch 20, batch 450, loss[loss=0.14, simple_loss=0.2152, pruned_loss=0.03243, over 7259.00 frames. ], tot_loss[loss=0.1511, simple_loss=0.2283, pruned_loss=0.03693, over 1291721.50 frames. ], batch size: 47, lr: 8.32e-03, grad_scale: 8.0 +2023-03-21 01:18:20,111 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 01:18:20,644 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 01:18:22,520 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 2.041e+02 2.494e+02 3.021e+02 4.369e+02, threshold=4.988e+02, percent-clipped=0.0 +2023-03-21 01:18:35,320 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.1332, 4.6256, 4.5840, 5.1408, 5.0408, 5.1428, 4.4122, 4.7913], + device='cuda:1'), covar=tensor([0.0770, 0.2415, 0.2086, 0.0937, 0.0773, 0.1102, 0.0856, 0.0833], + device='cuda:1'), in_proj_covar=tensor([0.0216, 0.0334, 0.0264, 0.0262, 0.0193, 0.0322, 0.0189, 0.0232], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:18:40,486 INFO [train.py:901] (1/2) Epoch 20, batch 500, loss[loss=0.14, simple_loss=0.2134, pruned_loss=0.03325, over 7154.00 frames. ], tot_loss[loss=0.1512, simple_loss=0.2282, pruned_loss=0.03706, over 1323314.32 frames. ], batch size: 41, lr: 8.32e-03, grad_scale: 8.0 +2023-03-21 01:18:52,588 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 01:18:54,074 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 01:18:54,155 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5234, 5.0391, 5.1443, 5.0551, 4.8868, 4.6191, 5.1758, 5.0027], + device='cuda:1'), covar=tensor([0.0414, 0.0347, 0.0331, 0.0420, 0.0319, 0.0322, 0.0275, 0.0441], + device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0218, 0.0164, 0.0160, 0.0130, 0.0198, 0.0170, 0.0130], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:18:54,645 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 01:18:57,150 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 01:18:57,500 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 +2023-03-21 01:19:02,204 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 01:19:06,577 INFO [train.py:901] (1/2) Epoch 20, batch 550, loss[loss=0.1366, simple_loss=0.2231, pruned_loss=0.02509, over 7348.00 frames. ], tot_loss[loss=0.1514, simple_loss=0.2284, pruned_loss=0.03725, over 1349645.64 frames. ], batch size: 63, lr: 8.32e-03, grad_scale: 8.0 +2023-03-21 01:19:14,155 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 01:19:14,608 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.132e+02 1.926e+02 2.263e+02 2.730e+02 5.297e+02, threshold=4.525e+02, percent-clipped=2.0 +2023-03-21 01:19:23,272 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 01:19:26,659 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 01:19:32,394 INFO [train.py:901] (1/2) Epoch 20, batch 600, loss[loss=0.158, simple_loss=0.2366, pruned_loss=0.03974, over 7265.00 frames. ], tot_loss[loss=0.1506, simple_loss=0.2277, pruned_loss=0.03681, over 1370832.76 frames. ], batch size: 70, lr: 8.31e-03, grad_scale: 8.0 +2023-03-21 01:19:33,904 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 01:19:50,692 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 01:19:58,384 INFO [train.py:901] (1/2) Epoch 20, batch 650, loss[loss=0.166, simple_loss=0.2458, pruned_loss=0.04315, over 7228.00 frames. ], tot_loss[loss=0.1511, simple_loss=0.2282, pruned_loss=0.03705, over 1388722.42 frames. ], batch size: 93, lr: 8.31e-03, grad_scale: 8.0 +2023-03-21 01:19:59,898 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 01:20:06,397 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.393e+02 2.135e+02 2.443e+02 2.990e+02 6.250e+02, threshold=4.886e+02, percent-clipped=3.0 +2023-03-21 01:20:08,173 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-03-21 01:20:16,656 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 01:20:24,917 INFO [train.py:901] (1/2) Epoch 20, batch 700, loss[loss=0.1192, simple_loss=0.193, pruned_loss=0.02269, over 7201.00 frames. ], tot_loss[loss=0.1519, simple_loss=0.229, pruned_loss=0.03746, over 1400433.45 frames. ], batch size: 39, lr: 8.30e-03, grad_scale: 8.0 +2023-03-21 01:20:25,467 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 01:20:49,351 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 01:20:49,853 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 01:20:50,347 INFO [train.py:901] (1/2) Epoch 20, batch 750, loss[loss=0.1339, simple_loss=0.2046, pruned_loss=0.03156, over 7194.00 frames. ], tot_loss[loss=0.1519, simple_loss=0.229, pruned_loss=0.03737, over 1411234.03 frames. ], batch size: 39, lr: 8.30e-03, grad_scale: 8.0 +2023-03-21 01:20:58,469 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.383e+02 1.864e+02 2.237e+02 2.756e+02 4.734e+02, threshold=4.474e+02, percent-clipped=0.0 +2023-03-21 01:21:00,261 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3813, 2.5811, 2.3029, 2.4226, 2.5167, 2.2910, 2.4805, 2.1672], + device='cuda:1'), covar=tensor([0.0869, 0.0800, 0.0744, 0.1015, 0.0485, 0.0840, 0.0850, 0.1523], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0045, 0.0053, 0.0047, 0.0047, 0.0047, 0.0047, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:21:04,604 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 01:21:09,617 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 01:21:16,109 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 01:21:16,596 INFO [train.py:901] (1/2) Epoch 20, batch 800, loss[loss=0.128, simple_loss=0.2038, pruned_loss=0.02603, over 7214.00 frames. ], tot_loss[loss=0.1512, simple_loss=0.2286, pruned_loss=0.03692, over 1419283.44 frames. ], batch size: 39, lr: 8.30e-03, grad_scale: 8.0 +2023-03-21 01:21:17,598 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 01:21:23,653 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5030, 2.7515, 3.2320, 3.4268, 3.4604, 3.3234, 3.2970, 3.1896], + device='cuda:1'), covar=tensor([0.0040, 0.0160, 0.0062, 0.0055, 0.0053, 0.0057, 0.0091, 0.0079], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0054, 0.0048, 0.0046, 0.0046, 0.0050, 0.0046, 0.0060], + device='cuda:1'), out_proj_covar=tensor([8.1828e-05, 1.2884e-04, 1.0687e-04, 9.7855e-05, 9.6831e-05, 1.0524e-04, + 1.0739e-04, 1.3029e-04], device='cuda:1') +2023-03-21 01:21:28,191 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 01:21:32,828 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2296, 4.3284, 3.6001, 3.6033, 3.6658, 2.4347, 2.1892, 4.3365], + device='cuda:1'), covar=tensor([0.0045, 0.0033, 0.0107, 0.0070, 0.0098, 0.0466, 0.0521, 0.0043], + device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0074, 0.0094, 0.0080, 0.0104, 0.0117, 0.0116, 0.0088], + device='cuda:1'), out_proj_covar=tensor([1.0502e-04, 9.8110e-05, 1.1990e-04, 1.0485e-04, 1.2997e-04, 1.4834e-04, + 1.4798e-04, 1.0814e-04], device='cuda:1') +2023-03-21 01:21:33,827 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3659, 3.9070, 3.7590, 4.4720, 4.2633, 4.3150, 3.8963, 3.8990], + device='cuda:1'), covar=tensor([0.1059, 0.2729, 0.2984, 0.0927, 0.0985, 0.1496, 0.0929, 0.1183], + device='cuda:1'), in_proj_covar=tensor([0.0218, 0.0326, 0.0265, 0.0257, 0.0190, 0.0321, 0.0190, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:21:41,933 INFO [train.py:901] (1/2) Epoch 20, batch 850, loss[loss=0.1529, simple_loss=0.2377, pruned_loss=0.03412, over 7313.00 frames. ], tot_loss[loss=0.1511, simple_loss=0.2286, pruned_loss=0.03678, over 1425199.14 frames. ], batch size: 83, lr: 8.29e-03, grad_scale: 8.0 +2023-03-21 01:21:43,078 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1984, 2.3735, 2.0309, 3.3929, 1.4817, 3.1329, 1.1877, 2.7605], + device='cuda:1'), covar=tensor([0.0122, 0.1060, 0.1858, 0.0098, 0.4134, 0.0142, 0.1154, 0.0325], + device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0261, 0.0283, 0.0175, 0.0270, 0.0190, 0.0256, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:21:47,580 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 01:21:48,076 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 01:21:50,593 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.337e+02 1.930e+02 2.357e+02 2.805e+02 7.911e+02, threshold=4.713e+02, percent-clipped=4.0 +2023-03-21 01:21:53,621 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 01:21:57,773 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 01:22:03,092 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2764, 3.6980, 3.8671, 3.9536, 3.7706, 3.8039, 4.0234, 3.5553], + device='cuda:1'), covar=tensor([0.0120, 0.0162, 0.0145, 0.0139, 0.0425, 0.0113, 0.0138, 0.0182], + device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0083, 0.0082, 0.0073, 0.0147, 0.0092, 0.0086, 0.0090], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:22:08,526 INFO [train.py:901] (1/2) Epoch 20, batch 900, loss[loss=0.1462, simple_loss=0.2253, pruned_loss=0.03361, over 7282.00 frames. ], tot_loss[loss=0.151, simple_loss=0.2283, pruned_loss=0.0368, over 1427402.17 frames. ], batch size: 47, lr: 8.29e-03, grad_scale: 16.0 +2023-03-21 01:22:34,245 INFO [train.py:901] (1/2) Epoch 20, batch 950, loss[loss=0.1447, simple_loss=0.2227, pruned_loss=0.03334, over 7326.00 frames. ], tot_loss[loss=0.1504, simple_loss=0.228, pruned_loss=0.03645, over 1432056.39 frames. ], batch size: 49, lr: 8.29e-03, grad_scale: 16.0 +2023-03-21 01:22:37,927 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 01:22:43,566 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.344e+02 2.000e+02 2.324e+02 2.740e+02 5.725e+02, threshold=4.649e+02, percent-clipped=1.0 +2023-03-21 01:22:57,550 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3051, 4.4986, 4.1753, 4.4481, 4.1037, 4.5493, 4.7143, 4.7629], + device='cuda:1'), covar=tensor([0.0188, 0.0133, 0.0189, 0.0165, 0.0311, 0.0158, 0.0240, 0.0195], + device='cuda:1'), in_proj_covar=tensor([0.0108, 0.0107, 0.0101, 0.0106, 0.0100, 0.0089, 0.0088, 0.0083], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:23:00,964 INFO [train.py:901] (1/2) Epoch 20, batch 1000, loss[loss=0.143, simple_loss=0.2296, pruned_loss=0.02822, over 7330.00 frames. ], tot_loss[loss=0.1512, simple_loss=0.2284, pruned_loss=0.03704, over 1431106.15 frames. ], batch size: 75, lr: 8.28e-03, grad_scale: 16.0 +2023-03-21 01:23:01,481 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 01:23:04,607 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54664.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 01:23:09,113 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8109, 3.0220, 2.5919, 2.6903, 2.8432, 2.6643, 2.8762, 2.5824], + device='cuda:1'), covar=tensor([0.0879, 0.0639, 0.1321, 0.2156, 0.0909, 0.0500, 0.1107, 0.3216], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0045, 0.0053, 0.0046, 0.0046, 0.0045, 0.0047, 0.0043], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:23:22,089 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 01:23:27,289 INFO [train.py:901] (1/2) Epoch 20, batch 1050, loss[loss=0.1624, simple_loss=0.2393, pruned_loss=0.04271, over 7251.00 frames. ], tot_loss[loss=0.151, simple_loss=0.2284, pruned_loss=0.03684, over 1431597.39 frames. ], batch size: 55, lr: 8.28e-03, grad_scale: 16.0 +2023-03-21 01:23:28,585 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 +2023-03-21 01:23:30,977 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2006, 1.3720, 1.2575, 1.3815, 1.3540, 1.3102, 1.1698, 0.9986], + device='cuda:1'), covar=tensor([0.0064, 0.0101, 0.0121, 0.0073, 0.0065, 0.0086, 0.0106, 0.0090], + device='cuda:1'), in_proj_covar=tensor([0.0025, 0.0024, 0.0024, 0.0024, 0.0023, 0.0024, 0.0025, 0.0032], + device='cuda:1'), out_proj_covar=tensor([2.8931e-05, 2.7913e-05, 2.7781e-05, 2.6599e-05, 2.6824e-05, 2.6777e-05, + 2.8995e-05, 3.6539e-05], device='cuda:1') +2023-03-21 01:23:35,312 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.346e+02 2.128e+02 2.433e+02 2.952e+02 5.627e+02, threshold=4.867e+02, percent-clipped=1.0 +2023-03-21 01:23:36,413 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54725.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 01:23:41,192 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-21 01:23:43,305 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 01:23:47,837 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 01:23:52,342 INFO [train.py:901] (1/2) Epoch 20, batch 1100, loss[loss=0.1332, simple_loss=0.2071, pruned_loss=0.02962, over 7253.00 frames. ], tot_loss[loss=0.151, simple_loss=0.2282, pruned_loss=0.03688, over 1434368.19 frames. ], batch size: 45, lr: 8.27e-03, grad_scale: 16.0 +2023-03-21 01:24:14,739 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1949, 0.9412, 1.3605, 1.6748, 1.3638, 1.6537, 1.1474, 1.5280], + device='cuda:1'), covar=tensor([0.1018, 0.2381, 0.0921, 0.0982, 0.1750, 0.0957, 0.1327, 0.1484], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0060, 0.0046, 0.0041, 0.0046, 0.0046, 0.0067, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 01:24:16,955 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 01:24:16,967 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 01:24:18,963 INFO [train.py:901] (1/2) Epoch 20, batch 1150, loss[loss=0.1764, simple_loss=0.2537, pruned_loss=0.04957, over 7349.00 frames. ], tot_loss[loss=0.1516, simple_loss=0.2285, pruned_loss=0.03734, over 1433743.51 frames. ], batch size: 63, lr: 8.27e-03, grad_scale: 16.0 +2023-03-21 01:24:26,905 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.285e+02 2.011e+02 2.444e+02 2.954e+02 6.699e+02, threshold=4.887e+02, percent-clipped=3.0 +2023-03-21 01:24:28,956 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 01:24:29,962 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 01:24:33,180 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9459, 2.1837, 2.6009, 2.1079, 2.5053, 2.0141, 1.9164, 1.5369], + device='cuda:1'), covar=tensor([0.0499, 0.0385, 0.0162, 0.0194, 0.0476, 0.0268, 0.0265, 0.0472], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0027, 0.0024, 0.0025, 0.0026, 0.0024, 0.0028, 0.0028], + device='cuda:1'), out_proj_covar=tensor([7.0977e-05, 6.9305e-05, 6.0952e-05, 6.2359e-05, 6.5970e-05, 6.3211e-05, + 6.9742e-05, 7.1514e-05], device='cuda:1') +2023-03-21 01:24:40,757 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8741, 2.1704, 1.5981, 2.6558, 2.2949, 2.4365, 1.9850, 2.3215], + device='cuda:1'), covar=tensor([0.1648, 0.0725, 0.2800, 0.0646, 0.0134, 0.0151, 0.0239, 0.0242], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0233, 0.0267, 0.0262, 0.0154, 0.0156, 0.0190, 0.0201], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:24:44,019 INFO [train.py:901] (1/2) Epoch 20, batch 1200, loss[loss=0.1494, simple_loss=0.2351, pruned_loss=0.03184, over 7256.00 frames. ], tot_loss[loss=0.1515, simple_loss=0.2287, pruned_loss=0.03714, over 1437190.72 frames. ], batch size: 89, lr: 8.27e-03, grad_scale: 16.0 +2023-03-21 01:24:46,667 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54862.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:24:49,324 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1291, 2.9649, 1.9632, 3.4579, 2.4141, 2.8127, 1.4870, 1.9286], + device='cuda:1'), covar=tensor([0.0383, 0.0718, 0.2574, 0.0637, 0.0467, 0.0534, 0.3335, 0.1887], + device='cuda:1'), in_proj_covar=tensor([0.0265, 0.0251, 0.0301, 0.0258, 0.0264, 0.0260, 0.0267, 0.0278], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 01:25:02,835 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 01:25:10,458 INFO [train.py:901] (1/2) Epoch 20, batch 1250, loss[loss=0.1435, simple_loss=0.2245, pruned_loss=0.03122, over 7314.00 frames. ], tot_loss[loss=0.1518, simple_loss=0.2287, pruned_loss=0.03748, over 1436113.03 frames. ], batch size: 83, lr: 8.26e-03, grad_scale: 16.0 +2023-03-21 01:25:18,346 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.491e+02 2.111e+02 2.421e+02 2.802e+02 5.340e+02, threshold=4.843e+02, percent-clipped=1.0 +2023-03-21 01:25:18,515 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54923.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 01:25:27,298 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 01:25:30,421 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8741, 3.8296, 3.1694, 3.2385, 2.9278, 2.2666, 1.6277, 3.8747], + device='cuda:1'), covar=tensor([0.0032, 0.0042, 0.0114, 0.0064, 0.0116, 0.0384, 0.0511, 0.0041], + device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0075, 0.0095, 0.0081, 0.0104, 0.0118, 0.0117, 0.0088], + device='cuda:1'), out_proj_covar=tensor([1.0591e-04, 9.9534e-05, 1.2100e-04, 1.0610e-04, 1.2925e-04, 1.4964e-04, + 1.4907e-04, 1.0884e-04], device='cuda:1') +2023-03-21 01:25:30,828 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 01:25:32,356 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 01:25:35,330 INFO [train.py:901] (1/2) Epoch 20, batch 1300, loss[loss=0.1684, simple_loss=0.2395, pruned_loss=0.04864, over 7307.00 frames. ], tot_loss[loss=0.1521, simple_loss=0.2292, pruned_loss=0.03748, over 1438486.26 frames. ], batch size: 86, lr: 8.26e-03, grad_scale: 16.0 +2023-03-21 01:25:55,473 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 01:25:57,965 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 01:26:01,813 INFO [train.py:901] (1/2) Epoch 20, batch 1350, loss[loss=0.1558, simple_loss=0.2264, pruned_loss=0.04263, over 7252.00 frames. ], tot_loss[loss=0.1522, simple_loss=0.2295, pruned_loss=0.03747, over 1441029.96 frames. ], batch size: 47, lr: 8.26e-03, grad_scale: 16.0 +2023-03-21 01:26:01,850 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 01:26:08,441 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55020.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 01:26:09,865 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.235e+02 1.994e+02 2.329e+02 2.781e+02 4.639e+02, threshold=4.658e+02, percent-clipped=0.0 +2023-03-21 01:26:12,867 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 01:26:28,033 INFO [train.py:901] (1/2) Epoch 20, batch 1400, loss[loss=0.1516, simple_loss=0.2355, pruned_loss=0.03383, over 7206.00 frames. ], tot_loss[loss=0.1517, simple_loss=0.2288, pruned_loss=0.03728, over 1440365.56 frames. ], batch size: 50, lr: 8.25e-03, grad_scale: 16.0 +2023-03-21 01:26:46,533 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 01:26:53,094 INFO [train.py:901] (1/2) Epoch 20, batch 1450, loss[loss=0.2141, simple_loss=0.2773, pruned_loss=0.07543, over 6767.00 frames. ], tot_loss[loss=0.1519, simple_loss=0.2291, pruned_loss=0.03738, over 1439840.67 frames. ], batch size: 106, lr: 8.25e-03, grad_scale: 16.0 +2023-03-21 01:26:56,649 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9653, 3.4145, 2.6619, 3.1328, 3.2775, 2.5725, 2.7228, 2.9644], + device='cuda:1'), covar=tensor([0.0967, 0.1027, 0.1263, 0.1317, 0.1236, 0.0942, 0.2898, 0.0944], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0046, 0.0054, 0.0047, 0.0046, 0.0047, 0.0048, 0.0043], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:27:00,980 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 1.941e+02 2.292e+02 2.919e+02 8.042e+02, threshold=4.584e+02, percent-clipped=3.0 +2023-03-21 01:27:02,661 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5504, 3.2230, 3.3386, 3.2533, 2.8525, 2.7200, 3.3691, 2.6526], + device='cuda:1'), covar=tensor([0.0300, 0.0293, 0.0377, 0.0392, 0.0483, 0.0590, 0.0506, 0.1185], + device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0331, 0.0266, 0.0352, 0.0307, 0.0303, 0.0337, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:27:10,719 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 01:27:14,663 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55148.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:27:19,146 INFO [train.py:901] (1/2) Epoch 20, batch 1500, loss[loss=0.1658, simple_loss=0.2379, pruned_loss=0.04684, over 7288.00 frames. ], tot_loss[loss=0.152, simple_loss=0.2293, pruned_loss=0.03735, over 1442398.08 frames. ], batch size: 66, lr: 8.24e-03, grad_scale: 16.0 +2023-03-21 01:27:19,801 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0240, 4.0267, 3.5971, 3.4603, 3.4431, 2.2687, 1.9597, 4.0944], + device='cuda:1'), covar=tensor([0.0036, 0.0032, 0.0079, 0.0058, 0.0075, 0.0417, 0.0468, 0.0037], + device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0074, 0.0094, 0.0081, 0.0104, 0.0116, 0.0116, 0.0088], + device='cuda:1'), out_proj_covar=tensor([1.0644e-04, 9.9276e-05, 1.2004e-04, 1.0633e-04, 1.2939e-04, 1.4826e-04, + 1.4857e-04, 1.0816e-04], device='cuda:1') +2023-03-21 01:27:26,867 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 01:27:43,876 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1042, 4.0285, 3.6585, 3.5412, 3.3914, 2.4929, 1.8675, 4.1589], + device='cuda:1'), covar=tensor([0.0042, 0.0064, 0.0085, 0.0055, 0.0096, 0.0411, 0.0503, 0.0046], + device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0075, 0.0095, 0.0081, 0.0105, 0.0117, 0.0117, 0.0088], + device='cuda:1'), out_proj_covar=tensor([1.0644e-04, 9.9630e-05, 1.2072e-04, 1.0672e-04, 1.3071e-04, 1.4920e-04, + 1.4861e-04, 1.0829e-04], device='cuda:1') +2023-03-21 01:27:44,731 INFO [train.py:901] (1/2) Epoch 20, batch 1550, loss[loss=0.1622, simple_loss=0.2387, pruned_loss=0.04287, over 7214.00 frames. ], tot_loss[loss=0.1514, simple_loss=0.2287, pruned_loss=0.03706, over 1442343.80 frames. ], batch size: 50, lr: 8.24e-03, grad_scale: 16.0 +2023-03-21 01:27:45,867 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55209.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:27:45,998 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.78 vs. limit=5.0 +2023-03-21 01:27:46,841 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55211.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:27:49,742 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 01:27:50,291 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55218.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 01:27:50,363 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8039, 3.0806, 2.5181, 2.7985, 2.8937, 2.5842, 2.7722, 2.7796], + device='cuda:1'), covar=tensor([0.0516, 0.1047, 0.1578, 0.1528, 0.0963, 0.1117, 0.1171, 0.1263], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0047, 0.0055, 0.0048, 0.0047, 0.0049, 0.0049, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:27:53,355 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 2.045e+02 2.381e+02 2.824e+02 4.578e+02, threshold=4.763e+02, percent-clipped=0.0 +2023-03-21 01:28:01,401 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6210, 2.3878, 2.1891, 2.0917, 2.2753, 2.1585, 2.0907, 1.6106], + device='cuda:1'), covar=tensor([0.0781, 0.0270, 0.0289, 0.0174, 0.0537, 0.0256, 0.0147, 0.0380], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0026, 0.0024, 0.0025, 0.0026, 0.0025, 0.0028, 0.0028], + device='cuda:1'), out_proj_covar=tensor([7.1208e-05, 6.8266e-05, 6.1161e-05, 6.1903e-05, 6.6669e-05, 6.3579e-05, + 6.8817e-05, 7.1383e-05], device='cuda:1') +2023-03-21 01:28:10,703 INFO [train.py:901] (1/2) Epoch 20, batch 1600, loss[loss=0.1796, simple_loss=0.2523, pruned_loss=0.05345, over 7284.00 frames. ], tot_loss[loss=0.1518, simple_loss=0.2293, pruned_loss=0.0372, over 1445120.47 frames. ], batch size: 77, lr: 8.24e-03, grad_scale: 16.0 +2023-03-21 01:28:18,434 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55272.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:28:21,827 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 01:28:22,326 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 01:28:25,363 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 01:28:26,511 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9906, 2.4030, 3.0700, 2.9642, 2.9908, 2.8018, 2.3262, 2.8196], + device='cuda:1'), covar=tensor([0.1119, 0.0543, 0.1049, 0.1405, 0.0805, 0.0783, 0.3072, 0.1599], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0053, 0.0039, 0.0039, 0.0039, 0.0036, 0.0055, 0.0042], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 01:28:34,468 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 01:28:35,908 INFO [train.py:901] (1/2) Epoch 20, batch 1650, loss[loss=0.1498, simple_loss=0.2199, pruned_loss=0.0398, over 7261.00 frames. ], tot_loss[loss=0.1517, simple_loss=0.2292, pruned_loss=0.03704, over 1445481.49 frames. ], batch size: 47, lr: 8.23e-03, grad_scale: 16.0 +2023-03-21 01:28:39,035 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 01:28:43,772 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55320.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 01:28:45,235 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.261e+02 2.014e+02 2.351e+02 2.724e+02 5.713e+02, threshold=4.703e+02, percent-clipped=3.0 +2023-03-21 01:28:47,324 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 01:29:02,360 INFO [train.py:901] (1/2) Epoch 20, batch 1700, loss[loss=0.1288, simple_loss=0.2101, pruned_loss=0.02373, over 7153.00 frames. ], tot_loss[loss=0.1512, simple_loss=0.229, pruned_loss=0.03672, over 1443156.47 frames. ], batch size: 41, lr: 8.23e-03, grad_scale: 16.0 +2023-03-21 01:29:02,648 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-03-21 01:29:03,914 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 01:29:07,942 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 01:29:07,992 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=55368.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 01:29:17,611 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 01:29:29,225 INFO [train.py:901] (1/2) Epoch 20, batch 1750, loss[loss=0.159, simple_loss=0.2353, pruned_loss=0.04132, over 7294.00 frames. ], tot_loss[loss=0.1518, simple_loss=0.2296, pruned_loss=0.03703, over 1443971.52 frames. ], batch size: 57, lr: 8.23e-03, grad_scale: 16.0 +2023-03-21 01:29:37,409 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.350e+02 1.989e+02 2.275e+02 2.798e+02 4.613e+02, threshold=4.551e+02, percent-clipped=0.0 +2023-03-21 01:29:42,501 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 01:29:43,536 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 01:29:47,166 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5529, 3.6427, 2.8091, 3.0230, 2.4015, 2.1778, 1.6398, 3.5429], + device='cuda:1'), covar=tensor([0.0066, 0.0047, 0.0190, 0.0106, 0.0209, 0.0485, 0.0606, 0.0061], + device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0074, 0.0094, 0.0081, 0.0103, 0.0116, 0.0116, 0.0086], + device='cuda:1'), out_proj_covar=tensor([1.0517e-04, 9.8750e-05, 1.1996e-04, 1.0670e-04, 1.2846e-04, 1.4743e-04, + 1.4722e-04, 1.0616e-04], device='cuda:1') +2023-03-21 01:29:54,412 INFO [train.py:901] (1/2) Epoch 20, batch 1800, loss[loss=0.1393, simple_loss=0.2173, pruned_loss=0.03064, over 7356.00 frames. ], tot_loss[loss=0.1516, simple_loss=0.2292, pruned_loss=0.03695, over 1438464.97 frames. ], batch size: 73, lr: 8.22e-03, grad_scale: 16.0 +2023-03-21 01:30:04,272 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 01:30:16,323 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6489, 2.7301, 2.3097, 3.9194, 1.7922, 3.5873, 1.5224, 3.0754], + device='cuda:1'), covar=tensor([0.0093, 0.0892, 0.1511, 0.0095, 0.3611, 0.0177, 0.1038, 0.0254], + device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0268, 0.0287, 0.0179, 0.0270, 0.0194, 0.0257, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:30:19,158 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 01:30:19,207 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55504.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:30:20,711 INFO [train.py:901] (1/2) Epoch 20, batch 1850, loss[loss=0.1495, simple_loss=0.2284, pruned_loss=0.03526, over 7348.00 frames. ], tot_loss[loss=0.1512, simple_loss=0.2285, pruned_loss=0.03689, over 1438648.81 frames. ], batch size: 73, lr: 8.22e-03, grad_scale: 16.0 +2023-03-21 01:30:26,468 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55518.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 01:30:28,889 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.452e+02 1.968e+02 2.509e+02 2.876e+02 6.441e+02, threshold=5.018e+02, percent-clipped=4.0 +2023-03-21 01:30:29,425 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 01:30:38,211 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 +2023-03-21 01:30:45,482 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 01:30:45,985 INFO [train.py:901] (1/2) Epoch 20, batch 1900, loss[loss=0.1473, simple_loss=0.226, pruned_loss=0.03431, over 7271.00 frames. ], tot_loss[loss=0.1511, simple_loss=0.2283, pruned_loss=0.03697, over 1438118.76 frames. ], batch size: 52, lr: 8.21e-03, grad_scale: 16.0 +2023-03-21 01:30:50,585 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=55566.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:30:51,110 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55567.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:31:11,469 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0278, 3.2901, 2.8253, 3.1550, 3.1320, 2.9577, 3.2755, 2.8796], + device='cuda:1'), covar=tensor([0.0800, 0.1062, 0.1163, 0.1110, 0.0942, 0.0515, 0.0535, 0.1369], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0047, 0.0055, 0.0047, 0.0048, 0.0049, 0.0048, 0.0045], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:31:11,849 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 01:31:12,338 INFO [train.py:901] (1/2) Epoch 20, batch 1950, loss[loss=0.1528, simple_loss=0.2337, pruned_loss=0.03596, over 7242.00 frames. ], tot_loss[loss=0.1512, simple_loss=0.2286, pruned_loss=0.0369, over 1439548.19 frames. ], batch size: 64, lr: 8.21e-03, grad_scale: 16.0 +2023-03-21 01:31:20,367 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.368e+02 2.150e+02 2.553e+02 3.108e+02 5.927e+02, threshold=5.105e+02, percent-clipped=4.0 +2023-03-21 01:31:21,442 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 01:31:25,908 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 01:31:26,404 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 01:31:35,575 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5936, 3.6456, 3.4990, 3.7207, 3.2950, 3.6081, 3.9239, 3.9405], + device='cuda:1'), covar=tensor([0.0236, 0.0189, 0.0222, 0.0178, 0.0354, 0.0362, 0.0268, 0.0206], + device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0112, 0.0104, 0.0110, 0.0103, 0.0093, 0.0092, 0.0088], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:31:38,084 INFO [train.py:901] (1/2) Epoch 20, batch 2000, loss[loss=0.1381, simple_loss=0.2212, pruned_loss=0.0275, over 7346.00 frames. ], tot_loss[loss=0.1515, simple_loss=0.229, pruned_loss=0.03693, over 1442054.01 frames. ], batch size: 54, lr: 8.21e-03, grad_scale: 16.0 +2023-03-21 01:31:44,807 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 01:31:54,886 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 01:32:03,481 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 01:32:03,972 INFO [train.py:901] (1/2) Epoch 20, batch 2050, loss[loss=0.1585, simple_loss=0.2331, pruned_loss=0.0419, over 7309.00 frames. ], tot_loss[loss=0.1515, simple_loss=0.2291, pruned_loss=0.03692, over 1441830.43 frames. ], batch size: 49, lr: 8.20e-03, grad_scale: 16.0 +2023-03-21 01:32:06,595 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0104, 2.5890, 1.6512, 3.2448, 3.0448, 3.0873, 2.4621, 2.6592], + device='cuda:1'), covar=tensor([0.2041, 0.1065, 0.3747, 0.0446, 0.0154, 0.0160, 0.0258, 0.0325], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0230, 0.0265, 0.0261, 0.0155, 0.0158, 0.0191, 0.0203], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:32:10,082 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7926, 2.6731, 2.9596, 2.3063, 2.4821, 2.3874, 2.2188, 1.5433], + device='cuda:1'), covar=tensor([0.0894, 0.0249, 0.0112, 0.0326, 0.0402, 0.0276, 0.0225, 0.0520], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0026, 0.0023, 0.0024, 0.0025, 0.0024, 0.0027, 0.0028], + device='cuda:1'), out_proj_covar=tensor([6.9559e-05, 6.7178e-05, 5.9742e-05, 6.1335e-05, 6.5157e-05, 6.2827e-05, + 6.7258e-05, 7.0138e-05], device='cuda:1') +2023-03-21 01:32:11,989 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.451e+02 1.903e+02 2.186e+02 2.766e+02 4.460e+02, threshold=4.371e+02, percent-clipped=0.0 +2023-03-21 01:32:25,702 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 01:32:26,038 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55748.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:32:30,261 INFO [train.py:901] (1/2) Epoch 20, batch 2100, loss[loss=0.1331, simple_loss=0.2118, pruned_loss=0.02717, over 7369.00 frames. ], tot_loss[loss=0.1514, simple_loss=0.229, pruned_loss=0.03696, over 1441426.85 frames. ], batch size: 51, lr: 8.20e-03, grad_scale: 16.0 +2023-03-21 01:32:36,777 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 01:32:39,759 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 01:32:45,981 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9149, 2.2000, 2.4510, 2.2356, 2.4372, 2.3426, 2.0969, 1.6308], + device='cuda:1'), covar=tensor([0.0405, 0.0341, 0.0167, 0.0123, 0.0362, 0.0208, 0.0222, 0.0360], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0026, 0.0024, 0.0024, 0.0026, 0.0024, 0.0027, 0.0028], + device='cuda:1'), out_proj_covar=tensor([6.9430e-05, 6.7564e-05, 6.0367e-05, 6.1422e-05, 6.5440e-05, 6.3240e-05, + 6.7062e-05, 7.0362e-05], device='cuda:1') +2023-03-21 01:32:47,999 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7853, 3.0813, 2.7122, 2.8806, 3.0022, 2.7679, 3.1416, 2.8118], + device='cuda:1'), covar=tensor([0.0632, 0.0682, 0.0978, 0.1560, 0.0964, 0.0712, 0.0794, 0.1228], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0046, 0.0055, 0.0048, 0.0048, 0.0049, 0.0048, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:32:54,238 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55804.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:32:55,643 INFO [train.py:901] (1/2) Epoch 20, batch 2150, loss[loss=0.1407, simple_loss=0.2224, pruned_loss=0.0295, over 7338.00 frames. ], tot_loss[loss=0.1512, simple_loss=0.2288, pruned_loss=0.03681, over 1441289.45 frames. ], batch size: 54, lr: 8.20e-03, grad_scale: 16.0 +2023-03-21 01:32:56,830 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55809.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:33:03,699 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.263e+02 2.025e+02 2.413e+02 3.099e+02 7.099e+02, threshold=4.826e+02, percent-clipped=3.0 +2023-03-21 01:33:17,508 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55848.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:33:19,355 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=55852.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:33:21,762 INFO [train.py:901] (1/2) Epoch 20, batch 2200, loss[loss=0.1514, simple_loss=0.2253, pruned_loss=0.03874, over 7267.00 frames. ], tot_loss[loss=0.1513, simple_loss=0.2291, pruned_loss=0.0367, over 1443470.10 frames. ], batch size: 70, lr: 8.19e-03, grad_scale: 8.0 +2023-03-21 01:33:24,089 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.67 vs. limit=5.0 +2023-03-21 01:33:25,274 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 01:33:26,870 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55867.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:33:27,376 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6195, 1.4821, 1.8665, 2.0916, 1.8563, 2.0711, 1.8180, 1.8279], + device='cuda:1'), covar=tensor([0.1847, 0.2340, 0.1237, 0.1023, 0.1941, 0.3260, 0.1756, 0.5230], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0061, 0.0045, 0.0043, 0.0045, 0.0047, 0.0066, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 01:33:28,864 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3059, 4.7814, 4.8214, 4.7843, 4.7157, 4.3939, 4.8611, 4.7143], + device='cuda:1'), covar=tensor([0.0434, 0.0352, 0.0307, 0.0400, 0.0263, 0.0300, 0.0283, 0.0413], + device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0219, 0.0164, 0.0163, 0.0130, 0.0200, 0.0173, 0.0130], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:33:47,066 INFO [train.py:901] (1/2) Epoch 20, batch 2250, loss[loss=0.1635, simple_loss=0.2372, pruned_loss=0.04497, over 7261.00 frames. ], tot_loss[loss=0.1507, simple_loss=0.2285, pruned_loss=0.03642, over 1443222.09 frames. ], batch size: 64, lr: 8.19e-03, grad_scale: 8.0 +2023-03-21 01:33:48,214 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55909.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:33:51,814 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=55915.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:33:56,851 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 2.068e+02 2.360e+02 2.879e+02 4.084e+02, threshold=4.720e+02, percent-clipped=0.0 +2023-03-21 01:33:59,933 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 01:34:00,474 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 01:34:12,254 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 01:34:13,750 INFO [train.py:901] (1/2) Epoch 20, batch 2300, loss[loss=0.1639, simple_loss=0.2435, pruned_loss=0.04213, over 7272.00 frames. ], tot_loss[loss=0.151, simple_loss=0.2291, pruned_loss=0.03643, over 1443530.33 frames. ], batch size: 68, lr: 8.19e-03, grad_scale: 8.0 +2023-03-21 01:34:23,522 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3049, 3.6943, 3.9965, 3.9424, 3.8811, 3.8234, 4.1595, 3.6088], + device='cuda:1'), covar=tensor([0.0133, 0.0189, 0.0127, 0.0154, 0.0381, 0.0130, 0.0134, 0.0216], + device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0083, 0.0081, 0.0072, 0.0144, 0.0092, 0.0086, 0.0090], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:34:30,180 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3064, 1.0053, 1.5984, 1.7644, 1.4817, 1.7103, 1.2853, 1.4978], + device='cuda:1'), covar=tensor([0.1808, 0.3638, 0.1240, 0.1092, 0.2055, 0.2470, 0.1488, 0.2765], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0061, 0.0046, 0.0043, 0.0045, 0.0047, 0.0065, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 01:34:30,715 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7270, 2.5838, 2.1323, 3.9005, 1.5300, 3.4503, 1.3491, 3.2470], + device='cuda:1'), covar=tensor([0.0109, 0.1023, 0.1745, 0.0116, 0.4372, 0.0127, 0.1451, 0.0339], + device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0259, 0.0277, 0.0176, 0.0262, 0.0187, 0.0252, 0.0220], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:34:44,523 INFO [train.py:901] (1/2) Epoch 20, batch 2350, loss[loss=0.1528, simple_loss=0.2315, pruned_loss=0.03704, over 7222.00 frames. ], tot_loss[loss=0.1501, simple_loss=0.2285, pruned_loss=0.03586, over 1443070.62 frames. ], batch size: 50, lr: 8.18e-03, grad_scale: 8.0 +2023-03-21 01:34:53,105 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.335e+02 2.001e+02 2.385e+02 2.972e+02 5.246e+02, threshold=4.769e+02, percent-clipped=3.0 +2023-03-21 01:35:01,932 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1420, 3.1790, 1.8272, 3.2858, 3.1162, 3.7743, 2.9772, 2.7189], + device='cuda:1'), covar=tensor([0.1898, 0.0587, 0.3842, 0.0399, 0.0094, 0.0104, 0.0219, 0.0166], + device='cuda:1'), in_proj_covar=tensor([0.0265, 0.0228, 0.0260, 0.0256, 0.0152, 0.0156, 0.0189, 0.0201], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:35:03,244 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 01:35:09,358 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 01:35:09,843 INFO [train.py:901] (1/2) Epoch 20, batch 2400, loss[loss=0.1233, simple_loss=0.2003, pruned_loss=0.0231, over 7174.00 frames. ], tot_loss[loss=0.15, simple_loss=0.2281, pruned_loss=0.03596, over 1440143.83 frames. ], batch size: 39, lr: 8.18e-03, grad_scale: 8.0 +2023-03-21 01:35:20,540 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 01:35:23,712 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 01:35:26,547 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-21 01:35:34,739 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56104.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:35:36,185 INFO [train.py:901] (1/2) Epoch 20, batch 2450, loss[loss=0.1652, simple_loss=0.2368, pruned_loss=0.04681, over 7257.00 frames. ], tot_loss[loss=0.1505, simple_loss=0.2283, pruned_loss=0.03637, over 1440578.82 frames. ], batch size: 47, lr: 8.17e-03, grad_scale: 8.0 +2023-03-21 01:35:37,366 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0978, 3.1528, 1.8226, 3.5516, 3.3421, 3.8260, 2.9916, 3.0849], + device='cuda:1'), covar=tensor([0.2077, 0.0753, 0.4202, 0.0343, 0.0113, 0.0124, 0.0209, 0.0206], + device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0225, 0.0257, 0.0251, 0.0151, 0.0153, 0.0188, 0.0200], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:35:40,595 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-03-21 01:35:44,666 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.431e+02 2.178e+02 2.434e+02 2.805e+02 5.134e+02, threshold=4.868e+02, percent-clipped=1.0 +2023-03-21 01:35:46,479 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-03-21 01:35:49,352 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6388, 2.3762, 3.0118, 2.6844, 2.8216, 2.5484, 2.1727, 2.7784], + device='cuda:1'), covar=tensor([0.2150, 0.0991, 0.0765, 0.1526, 0.0798, 0.1288, 0.2416, 0.1627], + device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0055, 0.0041, 0.0042, 0.0042, 0.0037, 0.0057, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:35:49,739 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 01:35:53,957 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-21 01:35:55,306 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56145.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:36:01,183 INFO [train.py:901] (1/2) Epoch 20, batch 2500, loss[loss=0.1445, simple_loss=0.2225, pruned_loss=0.03324, over 7302.00 frames. ], tot_loss[loss=0.1509, simple_loss=0.2288, pruned_loss=0.03649, over 1443314.49 frames. ], batch size: 49, lr: 8.17e-03, grad_scale: 8.0 +2023-03-21 01:36:15,799 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 01:36:26,501 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56204.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:36:27,617 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56206.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:36:27,970 INFO [train.py:901] (1/2) Epoch 20, batch 2550, loss[loss=0.1628, simple_loss=0.24, pruned_loss=0.04284, over 7258.00 frames. ], tot_loss[loss=0.1517, simple_loss=0.2293, pruned_loss=0.03702, over 1443205.09 frames. ], batch size: 47, lr: 8.17e-03, grad_scale: 8.0 +2023-03-21 01:36:36,413 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.460e+02 2.002e+02 2.403e+02 2.852e+02 5.182e+02, threshold=4.807e+02, percent-clipped=1.0 +2023-03-21 01:36:53,748 INFO [train.py:901] (1/2) Epoch 20, batch 2600, loss[loss=0.122, simple_loss=0.1907, pruned_loss=0.02666, over 6204.00 frames. ], tot_loss[loss=0.151, simple_loss=0.2289, pruned_loss=0.03662, over 1443955.19 frames. ], batch size: 27, lr: 8.16e-03, grad_scale: 8.0 +2023-03-21 01:37:04,242 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0171, 3.5699, 3.7797, 3.6641, 3.6303, 3.5281, 3.8711, 3.5420], + device='cuda:1'), covar=tensor([0.0122, 0.0204, 0.0110, 0.0171, 0.0378, 0.0146, 0.0139, 0.0153], + device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0083, 0.0082, 0.0072, 0.0146, 0.0092, 0.0086, 0.0089], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:37:18,990 INFO [train.py:901] (1/2) Epoch 20, batch 2650, loss[loss=0.128, simple_loss=0.2143, pruned_loss=0.02087, over 7348.00 frames. ], tot_loss[loss=0.1509, simple_loss=0.2286, pruned_loss=0.03662, over 1443178.20 frames. ], batch size: 61, lr: 8.16e-03, grad_scale: 8.0 +2023-03-21 01:37:21,094 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0983, 3.9889, 3.5025, 3.4441, 3.3003, 2.3180, 1.7911, 4.1253], + device='cuda:1'), covar=tensor([0.0033, 0.0041, 0.0082, 0.0052, 0.0092, 0.0392, 0.0492, 0.0034], + device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0074, 0.0094, 0.0079, 0.0103, 0.0117, 0.0116, 0.0086], + device='cuda:1'), out_proj_covar=tensor([1.0705e-04, 9.8391e-05, 1.1922e-04, 1.0457e-04, 1.2763e-04, 1.4896e-04, + 1.4806e-04, 1.0489e-04], device='cuda:1') +2023-03-21 01:37:27,401 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.210e+02 2.152e+02 2.521e+02 3.004e+02 6.613e+02, threshold=5.042e+02, percent-clipped=2.0 +2023-03-21 01:37:29,085 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5884, 2.9753, 3.5692, 3.1874, 2.8125, 2.8585, 3.4687, 2.6822], + device='cuda:1'), covar=tensor([0.0216, 0.0270, 0.0322, 0.0348, 0.0470, 0.0628, 0.0346, 0.1054], + device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0323, 0.0259, 0.0344, 0.0298, 0.0296, 0.0326, 0.0282], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:37:31,591 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5915, 2.7961, 2.5143, 2.6699, 2.7651, 2.4916, 2.7155, 2.5912], + device='cuda:1'), covar=tensor([0.0839, 0.1051, 0.0953, 0.0918, 0.0560, 0.0899, 0.2147, 0.1211], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0047, 0.0054, 0.0048, 0.0048, 0.0048, 0.0049, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:37:35,450 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2572, 3.5508, 3.8480, 3.8505, 3.7230, 3.6395, 3.9579, 3.4889], + device='cuda:1'), covar=tensor([0.0105, 0.0223, 0.0149, 0.0150, 0.0417, 0.0157, 0.0163, 0.0176], + device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0084, 0.0081, 0.0072, 0.0145, 0.0092, 0.0086, 0.0089], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:37:43,700 INFO [train.py:901] (1/2) Epoch 20, batch 2700, loss[loss=0.1549, simple_loss=0.2311, pruned_loss=0.03936, over 7308.00 frames. ], tot_loss[loss=0.1515, simple_loss=0.2291, pruned_loss=0.03698, over 1441928.73 frames. ], batch size: 49, lr: 8.16e-03, grad_scale: 8.0 +2023-03-21 01:37:48,844 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 +2023-03-21 01:38:07,006 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56404.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:38:08,412 INFO [train.py:901] (1/2) Epoch 20, batch 2750, loss[loss=0.1614, simple_loss=0.2448, pruned_loss=0.03897, over 7293.00 frames. ], tot_loss[loss=0.1516, simple_loss=0.229, pruned_loss=0.03706, over 1440673.09 frames. ], batch size: 66, lr: 8.15e-03, grad_scale: 8.0 +2023-03-21 01:38:08,540 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7280, 1.8266, 2.1894, 1.9732, 2.0328, 2.1037, 1.8534, 1.5619], + device='cuda:1'), covar=tensor([0.0465, 0.0728, 0.0320, 0.0189, 0.0570, 0.0372, 0.0290, 0.0281], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0026, 0.0025, 0.0025, 0.0026, 0.0024, 0.0027, 0.0028], + device='cuda:1'), out_proj_covar=tensor([6.9848e-05, 6.7507e-05, 6.1772e-05, 6.2730e-05, 6.6847e-05, 6.3055e-05, + 6.8360e-05, 7.0503e-05], device='cuda:1') +2023-03-21 01:38:14,402 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.02 vs. limit=5.0 +2023-03-21 01:38:16,489 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.333e+02 2.245e+02 2.531e+02 3.013e+02 6.042e+02, threshold=5.062e+02, percent-clipped=2.0 +2023-03-21 01:38:30,020 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=56452.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:38:32,364 INFO [train.py:901] (1/2) Epoch 20, batch 2800, loss[loss=0.1599, simple_loss=0.2419, pruned_loss=0.03896, over 7142.00 frames. ], tot_loss[loss=0.1527, simple_loss=0.2301, pruned_loss=0.03763, over 1442379.72 frames. ], batch size: 98, lr: 8.15e-03, grad_scale: 8.0 +2023-03-21 01:38:56,494 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 01:38:56,547 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 01:38:56,572 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 01:38:56,588 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 01:38:57,067 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 01:38:57,082 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 01:38:57,093 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 01:38:57,194 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 01:38:57,269 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 01:38:57,284 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 01:38:57,301 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 01:38:57,460 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 01:38:57,466 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 01:38:57,473 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 01:38:57,609 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 01:38:57,612 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 01:38:57,994 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 01:39:03,159 INFO [train.py:901] (1/2) Epoch 21, batch 0, loss[loss=0.1546, simple_loss=0.2282, pruned_loss=0.0405, over 7279.00 frames. ], tot_loss[loss=0.1546, simple_loss=0.2282, pruned_loss=0.0405, over 7279.00 frames. ], batch size: 47, lr: 7.96e-03, grad_scale: 8.0 +2023-03-21 01:39:03,159 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 01:39:28,649 INFO [train.py:935] (1/2) Epoch 21, validation: loss=0.1655, simple_loss=0.2529, pruned_loss=0.03902, over 1622729.00 frames. +2023-03-21 01:39:28,650 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 01:39:30,294 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5688, 5.1070, 5.2178, 5.1491, 4.9628, 4.6380, 5.2500, 5.0346], + device='cuda:1'), covar=tensor([0.0440, 0.0387, 0.0328, 0.0428, 0.0283, 0.0308, 0.0277, 0.0468], + device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0219, 0.0164, 0.0162, 0.0130, 0.0198, 0.0172, 0.0129], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:39:35,706 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 01:39:39,190 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 +2023-03-21 01:39:39,370 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56501.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:39:40,911 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56504.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:39:45,829 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 01:39:51,134 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56523.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 01:39:51,457 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.209e+02 1.941e+02 2.187e+02 2.550e+02 5.143e+02, threshold=4.375e+02, percent-clipped=1.0 +2023-03-21 01:39:53,988 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 01:39:55,005 INFO [train.py:901] (1/2) Epoch 21, batch 50, loss[loss=0.1575, simple_loss=0.2406, pruned_loss=0.0372, over 7295.00 frames. ], tot_loss[loss=0.1501, simple_loss=0.2271, pruned_loss=0.03653, over 324701.75 frames. ], batch size: 49, lr: 7.96e-03, grad_scale: 8.0 +2023-03-21 01:39:56,560 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 01:39:59,185 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 01:39:59,349 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0593, 2.9602, 2.0614, 3.5121, 2.2409, 2.8465, 1.7251, 1.9546], + device='cuda:1'), covar=tensor([0.0406, 0.0684, 0.2304, 0.0526, 0.0465, 0.0502, 0.2933, 0.1751], + device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0247, 0.0290, 0.0253, 0.0264, 0.0259, 0.0255, 0.0277], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 01:40:05,808 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=56552.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:40:08,886 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6392, 2.8681, 2.2876, 4.0388, 1.7808, 3.3278, 1.5608, 3.2726], + device='cuda:1'), covar=tensor([0.0109, 0.0785, 0.1541, 0.0097, 0.3596, 0.0111, 0.1022, 0.0373], + device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0260, 0.0282, 0.0176, 0.0263, 0.0189, 0.0251, 0.0219], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:40:15,309 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 01:40:15,760 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 01:40:20,320 INFO [train.py:901] (1/2) Epoch 21, batch 100, loss[loss=0.1471, simple_loss=0.2262, pruned_loss=0.03399, over 7353.00 frames. ], tot_loss[loss=0.1504, simple_loss=0.229, pruned_loss=0.03592, over 573475.33 frames. ], batch size: 63, lr: 7.95e-03, grad_scale: 8.0 +2023-03-21 01:40:21,983 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56584.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 01:40:36,932 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 +2023-03-21 01:40:43,551 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.343e+02 2.050e+02 2.363e+02 2.734e+02 5.404e+02, threshold=4.726e+02, percent-clipped=1.0 +2023-03-21 01:40:47,112 INFO [train.py:901] (1/2) Epoch 21, batch 150, loss[loss=0.1469, simple_loss=0.2278, pruned_loss=0.03299, over 7285.00 frames. ], tot_loss[loss=0.1494, simple_loss=0.2277, pruned_loss=0.03558, over 763995.97 frames. ], batch size: 66, lr: 7.95e-03, grad_scale: 8.0 +2023-03-21 01:40:52,243 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3388, 4.3629, 3.8739, 3.7453, 3.4971, 2.4028, 1.8119, 4.5073], + device='cuda:1'), covar=tensor([0.0036, 0.0069, 0.0066, 0.0055, 0.0079, 0.0420, 0.0508, 0.0033], + device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0076, 0.0096, 0.0082, 0.0107, 0.0121, 0.0120, 0.0087], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 01:40:52,934 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.54 vs. limit=5.0 +2023-03-21 01:40:55,904 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.61 vs. limit=5.0 +2023-03-21 01:40:56,696 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0515, 4.6181, 4.4889, 5.0618, 4.9139, 5.0397, 4.3769, 4.5265], + device='cuda:1'), covar=tensor([0.0844, 0.2306, 0.2174, 0.0973, 0.0859, 0.1059, 0.0707, 0.0944], + device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0334, 0.0265, 0.0264, 0.0194, 0.0324, 0.0193, 0.0234], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:41:06,699 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3733, 0.9126, 1.7084, 1.9611, 1.5574, 1.8494, 1.4488, 1.7251], + device='cuda:1'), covar=tensor([0.2579, 0.3507, 0.1370, 0.1524, 0.1754, 0.1635, 0.1271, 0.1124], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0060, 0.0046, 0.0041, 0.0044, 0.0047, 0.0065, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 01:41:12,629 INFO [train.py:901] (1/2) Epoch 21, batch 200, loss[loss=0.1397, simple_loss=0.2195, pruned_loss=0.02989, over 7320.00 frames. ], tot_loss[loss=0.1502, simple_loss=0.2283, pruned_loss=0.03608, over 914503.26 frames. ], batch size: 49, lr: 7.95e-03, grad_scale: 8.0 +2023-03-21 01:41:16,095 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 01:41:21,719 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 01:41:27,691 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 01:41:34,706 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.304e+02 1.965e+02 2.278e+02 2.575e+02 4.496e+02, threshold=4.557e+02, percent-clipped=0.0 +2023-03-21 01:41:38,258 INFO [train.py:901] (1/2) Epoch 21, batch 250, loss[loss=0.1459, simple_loss=0.2323, pruned_loss=0.02971, over 7344.00 frames. ], tot_loss[loss=0.1507, simple_loss=0.2288, pruned_loss=0.03628, over 1032009.33 frames. ], batch size: 54, lr: 7.94e-03, grad_scale: 8.0 +2023-03-21 01:41:40,751 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 01:41:54,008 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5590, 3.2492, 3.3662, 3.2871, 2.8027, 2.8708, 3.4914, 2.5968], + device='cuda:1'), covar=tensor([0.0336, 0.0365, 0.0366, 0.0362, 0.0519, 0.0694, 0.0575, 0.1271], + device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0329, 0.0267, 0.0352, 0.0305, 0.0302, 0.0334, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:42:01,810 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 01:42:02,168 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 +2023-03-21 01:42:03,823 INFO [train.py:901] (1/2) Epoch 21, batch 300, loss[loss=0.1694, simple_loss=0.2405, pruned_loss=0.04915, over 7314.00 frames. ], tot_loss[loss=0.1507, simple_loss=0.2287, pruned_loss=0.03637, over 1123751.41 frames. ], batch size: 59, lr: 7.94e-03, grad_scale: 8.0 +2023-03-21 01:42:10,982 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 01:42:14,920 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56801.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:42:15,654 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 +2023-03-21 01:42:15,913 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56803.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:42:26,233 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 1.973e+02 2.320e+02 2.778e+02 6.179e+02, threshold=4.640e+02, percent-clipped=2.0 +2023-03-21 01:42:29,722 INFO [train.py:901] (1/2) Epoch 21, batch 350, loss[loss=0.1115, simple_loss=0.1854, pruned_loss=0.01881, over 7218.00 frames. ], tot_loss[loss=0.1497, simple_loss=0.2273, pruned_loss=0.0361, over 1192150.66 frames. ], batch size: 39, lr: 7.93e-03, grad_scale: 8.0 +2023-03-21 01:42:39,452 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=56849.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:42:44,903 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 01:42:47,027 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56864.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:42:51,275 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.72 vs. limit=5.0 +2023-03-21 01:42:55,133 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56879.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 01:42:56,053 INFO [train.py:901] (1/2) Epoch 21, batch 400, loss[loss=0.1463, simple_loss=0.2185, pruned_loss=0.03701, over 7355.00 frames. ], tot_loss[loss=0.1485, simple_loss=0.2262, pruned_loss=0.03543, over 1247545.34 frames. ], batch size: 73, lr: 7.93e-03, grad_scale: 8.0 +2023-03-21 01:43:00,795 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0514, 2.7014, 2.0955, 3.6284, 2.3814, 2.7152, 1.6481, 2.1297], + device='cuda:1'), covar=tensor([0.0472, 0.0686, 0.2556, 0.0497, 0.0564, 0.0386, 0.3157, 0.1645], + device='cuda:1'), in_proj_covar=tensor([0.0266, 0.0252, 0.0298, 0.0255, 0.0268, 0.0260, 0.0260, 0.0283], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 01:43:01,804 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5666, 3.3825, 2.5465, 4.0873, 2.9777, 3.3097, 2.1801, 2.3485], + device='cuda:1'), covar=tensor([0.0340, 0.0519, 0.2494, 0.0415, 0.0459, 0.0532, 0.2821, 0.2063], + device='cuda:1'), in_proj_covar=tensor([0.0265, 0.0251, 0.0298, 0.0254, 0.0267, 0.0260, 0.0259, 0.0282], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 01:43:09,753 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56908.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:43:12,305 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6157, 1.0494, 1.9340, 2.0248, 1.7246, 2.0322, 1.6434, 1.9367], + device='cuda:1'), covar=tensor([0.1364, 0.2650, 0.1034, 0.2547, 0.1131, 0.3319, 0.1269, 0.1271], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0062, 0.0047, 0.0043, 0.0046, 0.0050, 0.0069, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 01:43:17,668 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.336e+02 2.058e+02 2.428e+02 2.833e+02 5.139e+02, threshold=4.856e+02, percent-clipped=2.0 +2023-03-21 01:43:21,704 INFO [train.py:901] (1/2) Epoch 21, batch 450, loss[loss=0.1609, simple_loss=0.2342, pruned_loss=0.0438, over 7325.00 frames. ], tot_loss[loss=0.1487, simple_loss=0.2265, pruned_loss=0.03546, over 1290280.25 frames. ], batch size: 59, lr: 7.93e-03, grad_scale: 8.0 +2023-03-21 01:43:22,845 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6338, 1.2107, 2.0533, 2.1639, 1.7957, 2.1480, 1.8572, 1.9298], + device='cuda:1'), covar=tensor([0.3787, 0.5432, 0.1113, 0.1736, 0.3422, 0.4204, 0.1697, 0.2575], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0063, 0.0047, 0.0044, 0.0047, 0.0050, 0.0069, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 01:43:25,268 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 01:43:25,756 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 01:43:41,430 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56969.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:43:47,241 INFO [train.py:901] (1/2) Epoch 21, batch 500, loss[loss=0.142, simple_loss=0.2166, pruned_loss=0.03374, over 7166.00 frames. ], tot_loss[loss=0.1495, simple_loss=0.2273, pruned_loss=0.03591, over 1323891.61 frames. ], batch size: 41, lr: 7.92e-03, grad_scale: 8.0 +2023-03-21 01:43:53,334 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56993.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:43:54,342 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56995.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:43:54,395 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8658, 2.5089, 1.6643, 2.8564, 2.9504, 2.9408, 2.3220, 2.6662], + device='cuda:1'), covar=tensor([0.2004, 0.0744, 0.3403, 0.0500, 0.0162, 0.0157, 0.0238, 0.0221], + device='cuda:1'), in_proj_covar=tensor([0.0266, 0.0230, 0.0264, 0.0256, 0.0157, 0.0156, 0.0190, 0.0201], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:43:57,516 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 01:43:58,999 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 01:43:59,442 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 01:44:01,986 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 01:44:06,360 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 01:44:09,357 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.406e+02 1.969e+02 2.416e+02 3.009e+02 5.673e+02, threshold=4.831e+02, percent-clipped=3.0 +2023-03-21 01:44:12,848 INFO [train.py:901] (1/2) Epoch 21, batch 550, loss[loss=0.1473, simple_loss=0.2232, pruned_loss=0.03569, over 7274.00 frames. ], tot_loss[loss=0.1501, simple_loss=0.2278, pruned_loss=0.03623, over 1348359.16 frames. ], batch size: 52, lr: 7.92e-03, grad_scale: 8.0 +2023-03-21 01:44:14,512 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6063, 5.0514, 5.1011, 5.0677, 4.8949, 4.6550, 5.1525, 4.9492], + device='cuda:1'), covar=tensor([0.0369, 0.0356, 0.0334, 0.0378, 0.0266, 0.0314, 0.0260, 0.0407], + device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0220, 0.0165, 0.0161, 0.0132, 0.0199, 0.0173, 0.0130], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:44:18,251 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-03-21 01:44:18,402 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 01:44:25,128 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57054.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:44:26,106 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57056.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:44:26,956 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 01:44:29,888 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 01:44:37,319 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 01:44:38,202 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-03-21 01:44:38,300 INFO [train.py:901] (1/2) Epoch 21, batch 600, loss[loss=0.1365, simple_loss=0.2188, pruned_loss=0.02709, over 7248.00 frames. ], tot_loss[loss=0.1496, simple_loss=0.2273, pruned_loss=0.03598, over 1368763.84 frames. ], batch size: 55, lr: 7.92e-03, grad_scale: 8.0 +2023-03-21 01:44:40,043 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 +2023-03-21 01:44:55,081 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 01:45:00,646 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.412e+02 2.016e+02 2.506e+02 2.951e+02 4.891e+02, threshold=5.012e+02, percent-clipped=1.0 +2023-03-21 01:45:04,272 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 01:45:04,753 INFO [train.py:901] (1/2) Epoch 21, batch 650, loss[loss=0.1479, simple_loss=0.2287, pruned_loss=0.03352, over 7136.00 frames. ], tot_loss[loss=0.1496, simple_loss=0.2272, pruned_loss=0.03602, over 1386402.84 frames. ], batch size: 98, lr: 7.91e-03, grad_scale: 8.0 +2023-03-21 01:45:18,860 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57159.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:45:20,704 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 01:45:24,985 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.46 vs. limit=5.0 +2023-03-21 01:45:28,912 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57179.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 01:45:29,333 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 01:45:29,794 INFO [train.py:901] (1/2) Epoch 21, batch 700, loss[loss=0.1505, simple_loss=0.232, pruned_loss=0.03444, over 7280.00 frames. ], tot_loss[loss=0.1489, simple_loss=0.2264, pruned_loss=0.03571, over 1398163.44 frames. ], batch size: 70, lr: 7.91e-03, grad_scale: 8.0 +2023-03-21 01:45:33,093 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57186.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:45:38,539 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57197.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:45:53,099 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 1.965e+02 2.359e+02 2.948e+02 5.669e+02, threshold=4.718e+02, percent-clipped=2.0 +2023-03-21 01:45:54,626 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 01:45:54,670 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=57227.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 01:45:55,577 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 01:45:56,527 INFO [train.py:901] (1/2) Epoch 21, batch 750, loss[loss=0.1615, simple_loss=0.239, pruned_loss=0.04202, over 7289.00 frames. ], tot_loss[loss=0.1489, simple_loss=0.2264, pruned_loss=0.03571, over 1407175.85 frames. ], batch size: 70, lr: 7.91e-03, grad_scale: 8.0 +2023-03-21 01:46:03,225 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57244.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:46:04,688 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57247.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:46:09,656 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 01:46:10,336 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57258.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 01:46:13,278 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57264.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:46:13,732 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 01:46:20,814 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 01:46:21,819 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 01:46:22,279 INFO [train.py:901] (1/2) Epoch 21, batch 800, loss[loss=0.1566, simple_loss=0.2396, pruned_loss=0.03683, over 7140.00 frames. ], tot_loss[loss=0.1493, simple_loss=0.2268, pruned_loss=0.03592, over 1413705.85 frames. ], batch size: 98, lr: 7.90e-03, grad_scale: 8.0 +2023-03-21 01:46:27,351 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9386, 2.7446, 3.1436, 2.8084, 3.2896, 2.8267, 2.5720, 3.1907], + device='cuda:1'), covar=tensor([0.1898, 0.0676, 0.1102, 0.1683, 0.0569, 0.1283, 0.1868, 0.1001], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0055, 0.0041, 0.0042, 0.0040, 0.0038, 0.0056, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:46:32,299 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 01:46:34,915 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57305.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:46:44,206 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.475e+02 2.042e+02 2.374e+02 2.907e+02 5.577e+02, threshold=4.749e+02, percent-clipped=3.0 +2023-03-21 01:46:47,724 INFO [train.py:901] (1/2) Epoch 21, batch 850, loss[loss=0.1591, simple_loss=0.2426, pruned_loss=0.03778, over 7222.00 frames. ], tot_loss[loss=0.1494, simple_loss=0.2268, pruned_loss=0.03598, over 1420389.81 frames. ], batch size: 93, lr: 7.90e-03, grad_scale: 8.0 +2023-03-21 01:46:50,744 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 01:46:50,749 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 01:46:56,298 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 01:46:56,870 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57349.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:46:57,884 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57351.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:46:59,806 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 01:47:12,277 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 +2023-03-21 01:47:13,396 INFO [train.py:901] (1/2) Epoch 21, batch 900, loss[loss=0.192, simple_loss=0.2628, pruned_loss=0.06062, over 6714.00 frames. ], tot_loss[loss=0.1496, simple_loss=0.2273, pruned_loss=0.03597, over 1426679.27 frames. ], batch size: 107, lr: 7.90e-03, grad_scale: 8.0 +2023-03-21 01:47:21,524 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 +2023-03-21 01:47:22,689 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57398.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:47:35,955 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.448e+02 2.097e+02 2.421e+02 2.780e+02 5.960e+02, threshold=4.843e+02, percent-clipped=2.0 +2023-03-21 01:47:39,059 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 01:47:39,537 INFO [train.py:901] (1/2) Epoch 21, batch 950, loss[loss=0.1545, simple_loss=0.2306, pruned_loss=0.03918, over 7251.00 frames. ], tot_loss[loss=0.1488, simple_loss=0.2265, pruned_loss=0.03553, over 1430284.78 frames. ], batch size: 64, lr: 7.89e-03, grad_scale: 8.0 +2023-03-21 01:47:54,289 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:47:54,330 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:47:55,826 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57462.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:48:03,869 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 01:48:05,874 INFO [train.py:901] (1/2) Epoch 21, batch 1000, loss[loss=0.152, simple_loss=0.232, pruned_loss=0.03597, over 7300.00 frames. ], tot_loss[loss=0.1486, simple_loss=0.2261, pruned_loss=0.03553, over 1431251.07 frames. ], batch size: 86, lr: 7.89e-03, grad_scale: 8.0 +2023-03-21 01:48:12,862 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57495.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:48:16,609 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-03-21 01:48:17,434 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3020, 1.5101, 1.2316, 1.3308, 1.4634, 1.3189, 1.1008, 0.8851], + device='cuda:1'), covar=tensor([0.0160, 0.0137, 0.0282, 0.0145, 0.0140, 0.0128, 0.0185, 0.0153], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0027, 0.0026, 0.0026, 0.0024, 0.0025, 0.0026, 0.0034], + device='cuda:1'), out_proj_covar=tensor([3.1612e-05, 3.0333e-05, 2.9622e-05, 2.9737e-05, 2.8265e-05, 2.8315e-05, + 3.0281e-05, 3.9108e-05], device='cuda:1') +2023-03-21 01:48:18,877 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=57507.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:48:23,897 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 01:48:27,042 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57523.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:48:27,396 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.303e+02 1.945e+02 2.283e+02 2.747e+02 4.323e+02, threshold=4.566e+02, percent-clipped=0.0 +2023-03-21 01:48:30,920 INFO [train.py:901] (1/2) Epoch 21, batch 1050, loss[loss=0.1579, simple_loss=0.237, pruned_loss=0.0394, over 7281.00 frames. ], tot_loss[loss=0.1487, simple_loss=0.226, pruned_loss=0.03565, over 1433280.61 frames. ], batch size: 70, lr: 7.89e-03, grad_scale: 8.0 +2023-03-21 01:48:36,935 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57542.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:48:42,466 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2345, 4.7169, 4.6719, 4.6426, 4.6479, 4.2834, 4.7785, 4.6047], + device='cuda:1'), covar=tensor([0.0547, 0.0398, 0.0483, 0.0571, 0.0309, 0.0415, 0.0345, 0.0477], + device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0217, 0.0164, 0.0160, 0.0131, 0.0198, 0.0174, 0.0129], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:48:42,474 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57553.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 01:48:44,036 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57556.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:48:44,443 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 01:48:46,694 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 +2023-03-21 01:48:48,679 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57564.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:48:49,599 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 01:48:56,971 INFO [train.py:901] (1/2) Epoch 21, batch 1100, loss[loss=0.1787, simple_loss=0.2483, pruned_loss=0.05454, over 7285.00 frames. ], tot_loss[loss=0.15, simple_loss=0.2274, pruned_loss=0.03628, over 1433597.35 frames. ], batch size: 66, lr: 7.88e-03, grad_scale: 8.0 +2023-03-21 01:49:07,085 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57600.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:49:13,263 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=57612.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:49:19,425 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 01:49:19,888 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.249e+02 1.979e+02 2.394e+02 2.879e+02 9.847e+02, threshold=4.789e+02, percent-clipped=4.0 +2023-03-21 01:49:19,934 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 01:49:22,649 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6140, 2.3587, 2.9699, 2.7147, 2.8698, 2.4584, 2.2021, 2.7036], + device='cuda:1'), covar=tensor([0.1901, 0.1059, 0.0959, 0.1423, 0.0697, 0.0997, 0.2604, 0.1404], + device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0056, 0.0042, 0.0042, 0.0040, 0.0038, 0.0056, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:49:23,549 INFO [train.py:901] (1/2) Epoch 21, batch 1150, loss[loss=0.1464, simple_loss=0.2302, pruned_loss=0.03131, over 7280.00 frames. ], tot_loss[loss=0.1492, simple_loss=0.2268, pruned_loss=0.03574, over 1436745.57 frames. ], batch size: 66, lr: 7.88e-03, grad_scale: 8.0 +2023-03-21 01:49:32,131 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 01:49:32,701 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57649.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:49:33,093 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 01:49:33,628 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57651.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:49:49,213 INFO [train.py:901] (1/2) Epoch 21, batch 1200, loss[loss=0.1565, simple_loss=0.2356, pruned_loss=0.03872, over 7352.00 frames. ], tot_loss[loss=0.1487, simple_loss=0.2264, pruned_loss=0.03545, over 1438169.15 frames. ], batch size: 73, lr: 7.88e-03, grad_scale: 8.0 +2023-03-21 01:49:57,356 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=57697.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:49:58,366 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=57699.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:50:04,857 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 01:50:11,323 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 1.984e+02 2.400e+02 2.943e+02 4.404e+02, threshold=4.799e+02, percent-clipped=0.0 +2023-03-21 01:50:14,858 INFO [train.py:901] (1/2) Epoch 21, batch 1250, loss[loss=0.1415, simple_loss=0.2153, pruned_loss=0.03388, over 7279.00 frames. ], tot_loss[loss=0.1488, simple_loss=0.2266, pruned_loss=0.03548, over 1441861.53 frames. ], batch size: 52, lr: 7.87e-03, grad_scale: 8.0 +2023-03-21 01:50:16,022 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57733.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:50:23,706 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4202, 0.9504, 1.7261, 1.8654, 1.6523, 1.7455, 1.5323, 1.7782], + device='cuda:1'), covar=tensor([0.1945, 0.3930, 0.0773, 0.1625, 0.1746, 0.1907, 0.1073, 0.2964], + device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0061, 0.0047, 0.0044, 0.0047, 0.0050, 0.0069, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 01:50:27,145 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57754.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:50:29,232 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 01:50:32,820 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 01:50:33,813 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 01:50:39,382 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3435, 2.3261, 2.0938, 3.5394, 1.5769, 3.2109, 1.4098, 3.1794], + device='cuda:1'), covar=tensor([0.0118, 0.1119, 0.1607, 0.0116, 0.3605, 0.0155, 0.1371, 0.0251], + device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0263, 0.0283, 0.0183, 0.0271, 0.0195, 0.0255, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:50:40,686 INFO [train.py:901] (1/2) Epoch 21, batch 1300, loss[loss=0.164, simple_loss=0.2355, pruned_loss=0.04628, over 7267.00 frames. ], tot_loss[loss=0.1488, simple_loss=0.2267, pruned_loss=0.03541, over 1443769.48 frames. ], batch size: 57, lr: 7.87e-03, grad_scale: 8.0 +2023-03-21 01:50:42,764 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2578, 2.3210, 2.2664, 2.1296, 2.1209, 1.8050, 1.7304, 1.5817], + device='cuda:1'), covar=tensor([0.0616, 0.0354, 0.0235, 0.0250, 0.0417, 0.0596, 0.0283, 0.0215], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0028, 0.0025, 0.0025, 0.0027, 0.0025, 0.0029, 0.0028], + device='cuda:1'), out_proj_covar=tensor([7.0145e-05, 7.0837e-05, 6.3140e-05, 6.3659e-05, 6.8065e-05, 6.5577e-05, + 7.1581e-05, 7.2067e-05], device='cuda:1') +2023-03-21 01:50:47,193 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57794.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:50:48,174 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6878, 3.8351, 3.6389, 3.8413, 3.4582, 3.7506, 4.1483, 4.1462], + device='cuda:1'), covar=tensor([0.0227, 0.0162, 0.0213, 0.0182, 0.0366, 0.0312, 0.0174, 0.0158], + device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0111, 0.0103, 0.0110, 0.0101, 0.0093, 0.0090, 0.0088], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:50:56,668 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 01:50:58,707 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 01:51:00,275 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57818.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:51:03,385 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 01:51:03,841 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.320e+02 1.899e+02 2.250e+02 2.724e+02 9.280e+02, threshold=4.499e+02, percent-clipped=3.0 +2023-03-21 01:51:07,387 INFO [train.py:901] (1/2) Epoch 21, batch 1350, loss[loss=0.1867, simple_loss=0.2661, pruned_loss=0.05359, over 6753.00 frames. ], tot_loss[loss=0.1489, simple_loss=0.2271, pruned_loss=0.03538, over 1443749.03 frames. ], batch size: 107, lr: 7.87e-03, grad_scale: 8.0 +2023-03-21 01:51:12,717 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 01:51:12,793 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57842.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:51:17,254 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57851.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:51:18,272 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57853.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 01:51:29,368 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8815, 3.8189, 3.2674, 3.3425, 2.8695, 2.0700, 1.7033, 3.9204], + device='cuda:1'), covar=tensor([0.0038, 0.0051, 0.0103, 0.0072, 0.0144, 0.0501, 0.0572, 0.0040], + device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0076, 0.0096, 0.0081, 0.0106, 0.0120, 0.0119, 0.0087], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 01:51:32,189 INFO [train.py:901] (1/2) Epoch 21, batch 1400, loss[loss=0.164, simple_loss=0.2412, pruned_loss=0.0434, over 7291.00 frames. ], tot_loss[loss=0.1488, simple_loss=0.2268, pruned_loss=0.03543, over 1442193.19 frames. ], batch size: 86, lr: 7.86e-03, grad_scale: 16.0 +2023-03-21 01:51:37,324 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=57890.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:51:42,473 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57900.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:51:42,899 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=57901.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:51:44,162 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 +2023-03-21 01:51:45,502 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 01:51:54,954 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 2.029e+02 2.540e+02 2.891e+02 4.630e+02, threshold=5.079e+02, percent-clipped=1.0 +2023-03-21 01:51:57,606 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0834, 3.9931, 3.5483, 3.4925, 3.1508, 2.2833, 1.7788, 4.1012], + device='cuda:1'), covar=tensor([0.0032, 0.0032, 0.0087, 0.0064, 0.0113, 0.0460, 0.0549, 0.0042], + device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0075, 0.0095, 0.0081, 0.0105, 0.0119, 0.0119, 0.0087], + device='cuda:1'), out_proj_covar=tensor([1.0729e-04, 9.9495e-05, 1.1994e-04, 1.0619e-04, 1.2977e-04, 1.5084e-04, + 1.5095e-04, 1.0624e-04], device='cuda:1') +2023-03-21 01:51:58,495 INFO [train.py:901] (1/2) Epoch 21, batch 1450, loss[loss=0.1249, simple_loss=0.196, pruned_loss=0.02689, over 7013.00 frames. ], tot_loss[loss=0.1482, simple_loss=0.226, pruned_loss=0.03517, over 1440578.94 frames. ], batch size: 35, lr: 7.86e-03, grad_scale: 16.0 +2023-03-21 01:52:01,105 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8788, 3.8145, 3.2182, 3.3309, 2.8909, 2.1191, 1.7408, 3.8998], + device='cuda:1'), covar=tensor([0.0035, 0.0044, 0.0106, 0.0075, 0.0122, 0.0492, 0.0530, 0.0043], + device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0075, 0.0095, 0.0081, 0.0105, 0.0119, 0.0119, 0.0087], + device='cuda:1'), out_proj_covar=tensor([1.0719e-04, 9.9479e-05, 1.1976e-04, 1.0611e-04, 1.2960e-04, 1.5059e-04, + 1.5061e-04, 1.0598e-04], device='cuda:1') +2023-03-21 01:52:06,949 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=57948.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:52:08,918 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 01:52:12,462 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0475, 4.6006, 4.5364, 5.0591, 4.9237, 5.0188, 4.5576, 4.5555], + device='cuda:1'), covar=tensor([0.0856, 0.2706, 0.2413, 0.1126, 0.0862, 0.1171, 0.0689, 0.1144], + device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0343, 0.0275, 0.0274, 0.0198, 0.0336, 0.0197, 0.0243], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:52:23,980 INFO [train.py:901] (1/2) Epoch 21, batch 1500, loss[loss=0.1414, simple_loss=0.2226, pruned_loss=0.03011, over 7366.00 frames. ], tot_loss[loss=0.1494, simple_loss=0.2275, pruned_loss=0.03567, over 1442599.57 frames. ], batch size: 65, lr: 7.86e-03, grad_scale: 16.0 +2023-03-21 01:52:25,505 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 01:52:45,682 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=6.92 vs. limit=5.0 +2023-03-21 01:52:46,377 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.254e+02 1.930e+02 2.338e+02 2.743e+02 1.049e+03, threshold=4.675e+02, percent-clipped=2.0 +2023-03-21 01:52:46,733 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-21 01:52:49,906 INFO [train.py:901] (1/2) Epoch 21, batch 1550, loss[loss=0.1458, simple_loss=0.2217, pruned_loss=0.03493, over 7258.00 frames. ], tot_loss[loss=0.1489, simple_loss=0.2267, pruned_loss=0.03552, over 1440143.70 frames. ], batch size: 70, lr: 7.85e-03, grad_scale: 16.0 +2023-03-21 01:52:49,928 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 01:53:01,640 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58054.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:53:15,770 INFO [train.py:901] (1/2) Epoch 21, batch 1600, loss[loss=0.1527, simple_loss=0.2305, pruned_loss=0.03741, over 7362.00 frames. ], tot_loss[loss=0.1496, simple_loss=0.2274, pruned_loss=0.03586, over 1440989.85 frames. ], batch size: 51, lr: 7.85e-03, grad_scale: 16.0 +2023-03-21 01:53:20,332 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 01:53:20,395 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58089.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:53:20,835 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 01:53:24,360 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 01:53:26,876 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=58102.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:53:33,362 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 01:53:35,005 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58118.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:53:37,440 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 01:53:37,911 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.450e+02 2.083e+02 2.533e+02 2.946e+02 4.318e+02, threshold=5.065e+02, percent-clipped=0.0 +2023-03-21 01:53:41,399 INFO [train.py:901] (1/2) Epoch 21, batch 1650, loss[loss=0.1778, simple_loss=0.2531, pruned_loss=0.0513, over 7295.00 frames. ], tot_loss[loss=0.1504, simple_loss=0.2284, pruned_loss=0.03619, over 1442439.10 frames. ], batch size: 57, lr: 7.85e-03, grad_scale: 16.0 +2023-03-21 01:53:43,012 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0180, 2.3994, 1.7765, 2.7568, 2.8566, 3.0928, 2.5810, 2.7739], + device='cuda:1'), covar=tensor([0.2225, 0.1106, 0.3374, 0.0722, 0.0166, 0.0157, 0.0277, 0.0393], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0235, 0.0269, 0.0262, 0.0163, 0.0161, 0.0191, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:53:45,346 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 01:53:45,682 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-21 01:53:46,438 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58141.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:53:47,951 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58144.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:53:52,004 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58151.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:53:59,536 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=58166.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:54:00,795 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 +2023-03-21 01:54:04,078 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 01:54:07,588 INFO [train.py:901] (1/2) Epoch 21, batch 1700, loss[loss=0.1338, simple_loss=0.2192, pruned_loss=0.02417, over 7290.00 frames. ], tot_loss[loss=0.1493, simple_loss=0.2269, pruned_loss=0.03585, over 1440569.25 frames. ], batch size: 68, lr: 7.84e-03, grad_scale: 16.0 +2023-03-21 01:54:08,121 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 01:54:11,252 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1245, 3.1507, 2.1300, 3.5776, 2.4787, 3.0546, 1.6751, 2.1054], + device='cuda:1'), covar=tensor([0.0378, 0.0814, 0.2259, 0.0695, 0.0444, 0.0851, 0.2998, 0.1663], + device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0255, 0.0298, 0.0261, 0.0268, 0.0261, 0.0258, 0.0280], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 01:54:16,682 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=58199.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:54:18,635 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58202.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:54:19,261 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.47 vs. limit=5.0 +2023-03-21 01:54:19,507 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 01:54:20,134 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58205.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:54:29,468 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.229e+02 2.007e+02 2.325e+02 2.996e+02 5.829e+02, threshold=4.650e+02, percent-clipped=1.0 +2023-03-21 01:54:32,994 INFO [train.py:901] (1/2) Epoch 21, batch 1750, loss[loss=0.1553, simple_loss=0.2309, pruned_loss=0.03982, over 7332.00 frames. ], tot_loss[loss=0.1491, simple_loss=0.2271, pruned_loss=0.03557, over 1439283.13 frames. ], batch size: 54, lr: 7.84e-03, grad_scale: 16.0 +2023-03-21 01:54:43,270 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 01:54:44,241 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 01:54:59,447 INFO [train.py:901] (1/2) Epoch 21, batch 1800, loss[loss=0.1499, simple_loss=0.2292, pruned_loss=0.03532, over 7293.00 frames. ], tot_loss[loss=0.1495, simple_loss=0.2276, pruned_loss=0.03574, over 1438751.98 frames. ], batch size: 86, lr: 7.84e-03, grad_scale: 16.0 +2023-03-21 01:55:06,573 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 01:55:20,264 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 01:55:21,703 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 2.040e+02 2.436e+02 2.888e+02 6.459e+02, threshold=4.872e+02, percent-clipped=2.0 +2023-03-21 01:55:25,268 INFO [train.py:901] (1/2) Epoch 21, batch 1850, loss[loss=0.1461, simple_loss=0.2238, pruned_loss=0.03418, over 7363.00 frames. ], tot_loss[loss=0.1495, simple_loss=0.2276, pruned_loss=0.03568, over 1438899.26 frames. ], batch size: 51, lr: 7.83e-03, grad_scale: 16.0 +2023-03-21 01:55:28,965 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6887, 5.3041, 5.3712, 5.2584, 5.0607, 4.6995, 5.3808, 5.1013], + device='cuda:1'), covar=tensor([0.0453, 0.0319, 0.0291, 0.0427, 0.0323, 0.0382, 0.0245, 0.0564], + device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0227, 0.0172, 0.0171, 0.0135, 0.0204, 0.0180, 0.0136], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 01:55:29,900 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 01:55:46,267 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 01:55:51,234 INFO [train.py:901] (1/2) Epoch 21, batch 1900, loss[loss=0.1505, simple_loss=0.2325, pruned_loss=0.03426, over 7322.00 frames. ], tot_loss[loss=0.149, simple_loss=0.227, pruned_loss=0.03552, over 1439257.24 frames. ], batch size: 59, lr: 7.83e-03, grad_scale: 16.0 +2023-03-21 01:55:55,324 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58389.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:56:11,940 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 01:56:13,962 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.177e+02 1.959e+02 2.260e+02 2.693e+02 5.838e+02, threshold=4.521e+02, percent-clipped=2.0 +2023-03-21 01:56:17,584 INFO [train.py:901] (1/2) Epoch 21, batch 1950, loss[loss=0.1536, simple_loss=0.2367, pruned_loss=0.03525, over 7257.00 frames. ], tot_loss[loss=0.1483, simple_loss=0.2262, pruned_loss=0.03523, over 1441262.32 frames. ], batch size: 55, lr: 7.83e-03, grad_scale: 16.0 +2023-03-21 01:56:17,854 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-03-21 01:56:21,168 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=58437.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:56:24,127 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 01:56:26,975 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-21 01:56:28,124 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 01:56:28,679 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 01:56:43,121 INFO [train.py:901] (1/2) Epoch 21, batch 2000, loss[loss=0.1459, simple_loss=0.2297, pruned_loss=0.03102, over 7280.00 frames. ], tot_loss[loss=0.1485, simple_loss=0.2264, pruned_loss=0.03533, over 1441301.00 frames. ], batch size: 70, lr: 7.82e-03, grad_scale: 16.0 +2023-03-21 01:56:45,634 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 01:56:51,771 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58497.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:56:53,302 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58500.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:56:56,795 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 01:57:04,236 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 01:57:05,204 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.357e+02 1.942e+02 2.312e+02 2.807e+02 4.995e+02, threshold=4.624e+02, percent-clipped=1.0 +2023-03-21 01:57:09,400 INFO [train.py:901] (1/2) Epoch 21, batch 2050, loss[loss=0.1507, simple_loss=0.2341, pruned_loss=0.03369, over 7261.00 frames. ], tot_loss[loss=0.1484, simple_loss=0.2263, pruned_loss=0.03521, over 1442036.23 frames. ], batch size: 89, lr: 7.82e-03, grad_scale: 16.0 +2023-03-21 01:57:10,494 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3010, 3.8981, 3.8682, 4.3468, 4.1879, 4.2726, 3.6249, 3.8365], + device='cuda:1'), covar=tensor([0.0901, 0.2577, 0.2148, 0.1122, 0.0942, 0.1323, 0.1034, 0.1127], + device='cuda:1'), in_proj_covar=tensor([0.0217, 0.0332, 0.0264, 0.0266, 0.0194, 0.0323, 0.0191, 0.0234], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:57:14,266 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-03-21 01:57:35,029 INFO [train.py:901] (1/2) Epoch 21, batch 2100, loss[loss=0.1796, simple_loss=0.257, pruned_loss=0.0511, over 6720.00 frames. ], tot_loss[loss=0.1487, simple_loss=0.2269, pruned_loss=0.03524, over 1440663.37 frames. ], batch size: 106, lr: 7.82e-03, grad_scale: 16.0 +2023-03-21 01:57:38,911 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 01:57:41,384 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 01:57:41,962 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9756, 4.1113, 3.9019, 4.0858, 3.6933, 4.0465, 4.3537, 4.4248], + device='cuda:1'), covar=tensor([0.0207, 0.0163, 0.0190, 0.0177, 0.0298, 0.0262, 0.0227, 0.0170], + device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0110, 0.0101, 0.0107, 0.0100, 0.0091, 0.0087, 0.0086], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 01:57:57,406 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.377e+02 1.997e+02 2.319e+02 3.007e+02 6.323e+02, threshold=4.638e+02, percent-clipped=6.0 +2023-03-21 01:58:00,914 INFO [train.py:901] (1/2) Epoch 21, batch 2150, loss[loss=0.174, simple_loss=0.2538, pruned_loss=0.04707, over 7120.00 frames. ], tot_loss[loss=0.1495, simple_loss=0.2277, pruned_loss=0.03565, over 1443386.34 frames. ], batch size: 98, lr: 7.81e-03, grad_scale: 16.0 +2023-03-21 01:58:22,882 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 +2023-03-21 01:58:26,532 INFO [train.py:901] (1/2) Epoch 21, batch 2200, loss[loss=0.1669, simple_loss=0.241, pruned_loss=0.04637, over 7330.00 frames. ], tot_loss[loss=0.1502, simple_loss=0.2282, pruned_loss=0.03604, over 1443933.39 frames. ], batch size: 75, lr: 7.81e-03, grad_scale: 16.0 +2023-03-21 01:58:26,673 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58681.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:58:26,914 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 +2023-03-21 01:58:27,528 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 01:58:48,510 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.237e+02 1.897e+02 2.320e+02 2.979e+02 4.171e+02, threshold=4.640e+02, percent-clipped=0.0 +2023-03-21 01:58:52,127 INFO [train.py:901] (1/2) Epoch 21, batch 2250, loss[loss=0.1302, simple_loss=0.1924, pruned_loss=0.034, over 6198.00 frames. ], tot_loss[loss=0.15, simple_loss=0.2283, pruned_loss=0.03582, over 1445155.89 frames. ], batch size: 27, lr: 7.81e-03, grad_scale: 16.0 +2023-03-21 01:58:53,297 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6715, 2.4805, 2.3771, 3.6949, 1.6794, 3.4116, 1.3987, 3.1923], + device='cuda:1'), covar=tensor([0.0102, 0.0976, 0.1307, 0.0113, 0.3623, 0.0157, 0.0999, 0.0313], + device='cuda:1'), in_proj_covar=tensor([0.0186, 0.0262, 0.0274, 0.0181, 0.0267, 0.0192, 0.0254, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 01:58:57,766 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58742.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:59:01,714 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 01:59:02,216 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 01:59:14,742 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 01:59:14,841 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58775.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:59:17,732 INFO [train.py:901] (1/2) Epoch 21, batch 2300, loss[loss=0.1536, simple_loss=0.2338, pruned_loss=0.03666, over 7289.00 frames. ], tot_loss[loss=0.1495, simple_loss=0.2279, pruned_loss=0.03557, over 1444789.38 frames. ], batch size: 80, lr: 7.80e-03, grad_scale: 16.0 +2023-03-21 01:59:26,488 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58797.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:59:28,406 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58800.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:59:40,445 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.260e+02 1.931e+02 2.266e+02 2.658e+02 4.867e+02, threshold=4.532e+02, percent-clipped=2.0 +2023-03-21 01:59:43,941 INFO [train.py:901] (1/2) Epoch 21, batch 2350, loss[loss=0.1547, simple_loss=0.2378, pruned_loss=0.03584, over 7336.00 frames. ], tot_loss[loss=0.1493, simple_loss=0.2277, pruned_loss=0.03546, over 1444024.71 frames. ], batch size: 61, lr: 7.80e-03, grad_scale: 16.0 +2023-03-21 01:59:46,618 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58836.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:59:51,618 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=58845.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:59:53,130 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=58848.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 01:59:55,210 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2483, 1.0148, 1.7230, 1.7154, 1.5738, 1.7035, 1.1914, 1.6871], + device='cuda:1'), covar=tensor([0.2206, 0.3193, 0.1060, 0.1381, 0.2137, 0.2313, 0.1812, 0.2065], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0061, 0.0047, 0.0044, 0.0046, 0.0049, 0.0070, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 02:00:01,561 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 02:00:02,659 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9147, 3.6654, 3.6913, 3.7390, 3.3319, 3.6015, 3.8881, 3.4349], + device='cuda:1'), covar=tensor([0.0222, 0.0184, 0.0160, 0.0185, 0.0735, 0.0167, 0.0195, 0.0205], + device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0084, 0.0084, 0.0075, 0.0150, 0.0094, 0.0089, 0.0092], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:00:08,776 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 02:00:10,246 INFO [train.py:901] (1/2) Epoch 21, batch 2400, loss[loss=0.1537, simple_loss=0.2363, pruned_loss=0.03558, over 7346.00 frames. ], tot_loss[loss=0.1493, simple_loss=0.2276, pruned_loss=0.03548, over 1445006.28 frames. ], batch size: 73, lr: 7.80e-03, grad_scale: 16.0 +2023-03-21 02:00:11,806 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0255, 4.5232, 4.5277, 4.4479, 4.4602, 4.0430, 4.5646, 4.3901], + device='cuda:1'), covar=tensor([0.0563, 0.0443, 0.0423, 0.0558, 0.0343, 0.0445, 0.0389, 0.0507], + device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0229, 0.0172, 0.0172, 0.0137, 0.0206, 0.0184, 0.0136], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 02:00:18,752 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 02:00:21,015 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.78 vs. limit=5.0 +2023-03-21 02:00:21,756 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 02:00:22,861 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58906.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:00:30,849 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0967, 2.6276, 1.9055, 2.8293, 2.9495, 3.1930, 2.6599, 2.3575], + device='cuda:1'), covar=tensor([0.1784, 0.0847, 0.3376, 0.0603, 0.0159, 0.0129, 0.0257, 0.0303], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0237, 0.0267, 0.0262, 0.0165, 0.0163, 0.0195, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:00:31,682 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.319e+02 1.956e+02 2.269e+02 2.844e+02 5.396e+02, threshold=4.539e+02, percent-clipped=1.0 +2023-03-21 02:00:35,224 INFO [train.py:901] (1/2) Epoch 21, batch 2450, loss[loss=0.1475, simple_loss=0.2278, pruned_loss=0.03366, over 7284.00 frames. ], tot_loss[loss=0.1494, simple_loss=0.2275, pruned_loss=0.03563, over 1441830.13 frames. ], batch size: 66, lr: 7.79e-03, grad_scale: 16.0 +2023-03-21 02:00:47,955 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 02:00:53,552 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 +2023-03-21 02:00:54,791 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58967.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:00:58,366 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3932, 3.2060, 3.2439, 3.1502, 2.8198, 2.7793, 3.5214, 2.6092], + device='cuda:1'), covar=tensor([0.0341, 0.0355, 0.0501, 0.0435, 0.0556, 0.0729, 0.0434, 0.1310], + device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0326, 0.0264, 0.0350, 0.0301, 0.0296, 0.0333, 0.0281], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:01:01,648 INFO [train.py:901] (1/2) Epoch 21, batch 2500, loss[loss=0.1666, simple_loss=0.242, pruned_loss=0.04558, over 7333.00 frames. ], tot_loss[loss=0.1491, simple_loss=0.2271, pruned_loss=0.03553, over 1443246.62 frames. ], batch size: 61, lr: 7.79e-03, grad_scale: 16.0 +2023-03-21 02:01:14,054 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 02:01:24,727 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.296e+02 1.969e+02 2.261e+02 3.016e+02 6.580e+02, threshold=4.523e+02, percent-clipped=4.0 +2023-03-21 02:01:27,648 INFO [train.py:901] (1/2) Epoch 21, batch 2550, loss[loss=0.1451, simple_loss=0.2252, pruned_loss=0.03255, over 7320.00 frames. ], tot_loss[loss=0.1492, simple_loss=0.2269, pruned_loss=0.03577, over 1440645.27 frames. ], batch size: 83, lr: 7.79e-03, grad_scale: 8.0 +2023-03-21 02:01:30,768 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59037.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:01:41,661 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7066, 2.1437, 1.6209, 2.6070, 2.7379, 2.7937, 2.1286, 2.2076], + device='cuda:1'), covar=tensor([0.2057, 0.0988, 0.3507, 0.0739, 0.0218, 0.0146, 0.0251, 0.0324], + device='cuda:1'), in_proj_covar=tensor([0.0266, 0.0232, 0.0260, 0.0258, 0.0162, 0.0161, 0.0190, 0.0205], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:01:47,091 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6524, 4.1955, 4.1032, 4.6574, 4.5614, 4.6155, 4.1357, 4.1444], + device='cuda:1'), covar=tensor([0.0818, 0.2372, 0.2162, 0.1026, 0.0718, 0.1105, 0.0732, 0.1149], + device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0337, 0.0268, 0.0274, 0.0200, 0.0331, 0.0194, 0.0239], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:01:53,456 INFO [train.py:901] (1/2) Epoch 21, batch 2600, loss[loss=0.1234, simple_loss=0.2061, pruned_loss=0.02037, over 7331.00 frames. ], tot_loss[loss=0.1487, simple_loss=0.2266, pruned_loss=0.03536, over 1442548.69 frames. ], batch size: 44, lr: 7.78e-03, grad_scale: 8.0 +2023-03-21 02:02:03,309 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 +2023-03-21 02:02:15,430 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.271e+02 1.926e+02 2.254e+02 2.670e+02 6.116e+02, threshold=4.508e+02, percent-clipped=2.0 +2023-03-21 02:02:18,493 INFO [train.py:901] (1/2) Epoch 21, batch 2650, loss[loss=0.1507, simple_loss=0.2289, pruned_loss=0.0362, over 7257.00 frames. ], tot_loss[loss=0.148, simple_loss=0.226, pruned_loss=0.03505, over 1441658.73 frames. ], batch size: 55, lr: 7.78e-03, grad_scale: 8.0 +2023-03-21 02:02:18,565 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59131.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:02:34,708 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59164.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:02:42,791 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 +2023-03-21 02:02:43,361 INFO [train.py:901] (1/2) Epoch 21, batch 2700, loss[loss=0.1442, simple_loss=0.2301, pruned_loss=0.02916, over 7251.00 frames. ], tot_loss[loss=0.1484, simple_loss=0.2265, pruned_loss=0.03512, over 1442608.79 frames. ], batch size: 64, lr: 7.78e-03, grad_scale: 8.0 +2023-03-21 02:03:00,998 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3699, 1.0433, 1.8950, 1.9129, 1.8221, 1.9699, 1.4858, 1.7878], + device='cuda:1'), covar=tensor([0.1659, 0.3281, 0.1013, 0.0939, 0.1408, 0.1355, 0.1676, 0.3035], + device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0063, 0.0049, 0.0045, 0.0049, 0.0051, 0.0072, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:03:04,670 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-21 02:03:05,337 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 1.893e+02 2.333e+02 2.936e+02 8.243e+02, threshold=4.666e+02, percent-clipped=3.0 +2023-03-21 02:03:05,492 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59225.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:03:08,325 INFO [train.py:901] (1/2) Epoch 21, batch 2750, loss[loss=0.1673, simple_loss=0.2512, pruned_loss=0.04168, over 7264.00 frames. ], tot_loss[loss=0.1487, simple_loss=0.2266, pruned_loss=0.03538, over 1443112.96 frames. ], batch size: 55, lr: 7.77e-03, grad_scale: 8.0 +2023-03-21 02:03:18,014 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-21 02:03:23,591 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59262.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:03:33,222 INFO [train.py:901] (1/2) Epoch 21, batch 2800, loss[loss=0.1526, simple_loss=0.2453, pruned_loss=0.02991, over 7253.00 frames. ], tot_loss[loss=0.1487, simple_loss=0.2267, pruned_loss=0.03536, over 1444583.13 frames. ], batch size: 55, lr: 7.77e-03, grad_scale: 8.0 +2023-03-21 02:03:59,485 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 02:03:59,539 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 02:03:59,568 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 02:03:59,585 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 02:03:59,754 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 02:03:59,769 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 02:03:59,780 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 02:03:59,879 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 02:03:59,954 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 02:03:59,969 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 02:03:59,986 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 02:04:00,484 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 02:04:00,490 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 02:04:00,497 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 02:04:00,634 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 02:04:00,636 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 02:04:00,694 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 02:04:10,091 INFO [train.py:901] (1/2) Epoch 22, batch 0, loss[loss=0.1464, simple_loss=0.227, pruned_loss=0.0329, over 7261.00 frames. ], tot_loss[loss=0.1464, simple_loss=0.227, pruned_loss=0.0329, over 7261.00 frames. ], batch size: 64, lr: 7.60e-03, grad_scale: 8.0 +2023-03-21 02:04:10,091 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 02:04:35,415 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0435, 4.1741, 4.0764, 4.2194, 3.7797, 4.0223, 4.4667, 4.4429], + device='cuda:1'), covar=tensor([0.0097, 0.0102, 0.0103, 0.0176, 0.0343, 0.0233, 0.0198, 0.0231], + device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0115, 0.0107, 0.0112, 0.0104, 0.0095, 0.0090, 0.0091], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:04:36,419 INFO [train.py:935] (1/2) Epoch 22, validation: loss=0.1648, simple_loss=0.253, pruned_loss=0.03832, over 1622729.00 frames. +2023-03-21 02:04:36,420 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 02:04:43,497 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 02:04:46,409 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.332e+02 1.893e+02 2.208e+02 2.728e+02 4.951e+02, threshold=4.416e+02, percent-clipped=1.0 +2023-03-21 02:04:52,392 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59337.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:04:52,874 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.8090, 5.3484, 5.4130, 5.3006, 5.0512, 4.8008, 5.4099, 5.1554], + device='cuda:1'), covar=tensor([0.0423, 0.0349, 0.0296, 0.0465, 0.0314, 0.0361, 0.0301, 0.0452], + device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0225, 0.0170, 0.0171, 0.0135, 0.0204, 0.0180, 0.0135], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 02:04:53,333 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 02:04:53,909 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8733, 3.8157, 3.0182, 3.3894, 3.0036, 2.0706, 1.6951, 3.9659], + device='cuda:1'), covar=tensor([0.0048, 0.0052, 0.0128, 0.0064, 0.0128, 0.0484, 0.0551, 0.0045], + device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0077, 0.0096, 0.0081, 0.0108, 0.0121, 0.0119, 0.0088], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 02:05:00,431 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 02:05:01,391 INFO [train.py:901] (1/2) Epoch 22, batch 50, loss[loss=0.1574, simple_loss=0.2277, pruned_loss=0.04357, over 7309.00 frames. ], tot_loss[loss=0.1471, simple_loss=0.2225, pruned_loss=0.03585, over 322280.63 frames. ], batch size: 59, lr: 7.59e-03, grad_scale: 8.0 +2023-03-21 02:05:02,422 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 02:05:05,493 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 02:05:13,953 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 +2023-03-21 02:05:16,771 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=59385.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:05:16,848 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7842, 2.2411, 2.8287, 2.6731, 2.6519, 2.5254, 2.2438, 2.6820], + device='cuda:1'), covar=tensor([0.0838, 0.0857, 0.0784, 0.0960, 0.0781, 0.1217, 0.2290, 0.1256], + device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0057, 0.0044, 0.0042, 0.0042, 0.0040, 0.0059, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:05:21,294 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4511, 3.2818, 3.1875, 3.1957, 2.7709, 2.8829, 3.3060, 2.6702], + device='cuda:1'), covar=tensor([0.0494, 0.0411, 0.0452, 0.0574, 0.0720, 0.0876, 0.0537, 0.1582], + device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0335, 0.0268, 0.0359, 0.0308, 0.0303, 0.0342, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:05:24,181 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 02:05:24,681 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 02:05:28,579 INFO [train.py:901] (1/2) Epoch 22, batch 100, loss[loss=0.1547, simple_loss=0.2383, pruned_loss=0.03552, over 7282.00 frames. ], tot_loss[loss=0.1483, simple_loss=0.2252, pruned_loss=0.03567, over 568588.02 frames. ], batch size: 86, lr: 7.59e-03, grad_scale: 8.0 +2023-03-21 02:05:38,494 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.392e+02 1.976e+02 2.370e+02 2.738e+02 5.901e+02, threshold=4.739e+02, percent-clipped=2.0 +2023-03-21 02:05:41,732 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59431.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:05:53,650 INFO [train.py:901] (1/2) Epoch 22, batch 150, loss[loss=0.1434, simple_loss=0.2228, pruned_loss=0.032, over 7274.00 frames. ], tot_loss[loss=0.1486, simple_loss=0.2254, pruned_loss=0.0359, over 761598.27 frames. ], batch size: 66, lr: 7.59e-03, grad_scale: 8.0 +2023-03-21 02:05:55,309 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7776, 2.3957, 2.9023, 2.6061, 2.6771, 2.3212, 1.8559, 1.8203], + device='cuda:1'), covar=tensor([0.0590, 0.0627, 0.0157, 0.0196, 0.0431, 0.0387, 0.0253, 0.0302], + device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0028, 0.0026, 0.0026, 0.0028, 0.0027, 0.0031, 0.0030], + device='cuda:1'), out_proj_covar=tensor([7.4877e-05, 7.3483e-05, 6.7023e-05, 6.6254e-05, 7.2206e-05, 6.9799e-05, + 7.5320e-05, 7.5607e-05], device='cuda:1') +2023-03-21 02:06:06,371 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=59479.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:06:13,156 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=7.09 vs. limit=5.0 +2023-03-21 02:06:17,970 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59502.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:06:19,296 INFO [train.py:901] (1/2) Epoch 22, batch 200, loss[loss=0.1557, simple_loss=0.2341, pruned_loss=0.03864, over 7274.00 frames. ], tot_loss[loss=0.1489, simple_loss=0.2259, pruned_loss=0.03596, over 912590.70 frames. ], batch size: 77, lr: 7.58e-03, grad_scale: 8.0 +2023-03-21 02:06:24,843 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 02:06:26,903 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59520.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:06:29,273 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.401e+02 1.757e+02 2.128e+02 2.718e+02 7.175e+02, threshold=4.257e+02, percent-clipped=2.0 +2023-03-21 02:06:29,308 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 02:06:35,276 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 02:06:44,167 INFO [train.py:901] (1/2) Epoch 22, batch 250, loss[loss=0.1694, simple_loss=0.2403, pruned_loss=0.04929, over 7366.00 frames. ], tot_loss[loss=0.1488, simple_loss=0.2257, pruned_loss=0.03595, over 1033198.32 frames. ], batch size: 51, lr: 7.58e-03, grad_scale: 8.0 +2023-03-21 02:06:48,277 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 02:06:48,360 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59562.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:06:49,520 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59563.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:07:09,199 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 02:07:10,670 INFO [train.py:901] (1/2) Epoch 22, batch 300, loss[loss=0.1571, simple_loss=0.2317, pruned_loss=0.04124, over 7301.00 frames. ], tot_loss[loss=0.1496, simple_loss=0.227, pruned_loss=0.03612, over 1125114.82 frames. ], batch size: 49, lr: 7.58e-03, grad_scale: 8.0 +2023-03-21 02:07:13,235 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=59610.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:07:17,250 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 02:07:20,811 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.341e+02 1.899e+02 2.338e+02 2.742e+02 5.758e+02, threshold=4.677e+02, percent-clipped=3.0 +2023-03-21 02:07:37,263 INFO [train.py:901] (1/2) Epoch 22, batch 350, loss[loss=0.1266, simple_loss=0.2055, pruned_loss=0.02387, over 7353.00 frames. ], tot_loss[loss=0.1486, simple_loss=0.2261, pruned_loss=0.03559, over 1193934.40 frames. ], batch size: 44, lr: 7.57e-03, grad_scale: 8.0 +2023-03-21 02:07:51,024 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6581, 2.1935, 2.1836, 2.0319, 2.1785, 2.0317, 1.6044, 1.5279], + device='cuda:1'), covar=tensor([0.0529, 0.0298, 0.0166, 0.0175, 0.0360, 0.0336, 0.0306, 0.0266], + device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0028, 0.0026, 0.0026, 0.0028, 0.0027, 0.0030, 0.0029], + device='cuda:1'), out_proj_covar=tensor([7.4634e-05, 7.3537e-05, 6.7020e-05, 6.6491e-05, 7.1874e-05, 6.9484e-05, + 7.5102e-05, 7.5235e-05], device='cuda:1') +2023-03-21 02:07:53,482 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 02:08:02,418 INFO [train.py:901] (1/2) Epoch 22, batch 400, loss[loss=0.143, simple_loss=0.2206, pruned_loss=0.03269, over 7293.00 frames. ], tot_loss[loss=0.1488, simple_loss=0.2264, pruned_loss=0.03562, over 1249043.49 frames. ], batch size: 68, lr: 7.57e-03, grad_scale: 8.0 +2023-03-21 02:08:10,107 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5973, 4.2133, 4.0099, 4.6248, 4.4903, 4.5364, 3.9307, 4.0450], + device='cuda:1'), covar=tensor([0.0792, 0.2113, 0.2319, 0.0933, 0.0841, 0.1077, 0.0741, 0.1175], + device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0333, 0.0266, 0.0266, 0.0199, 0.0326, 0.0193, 0.0235], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:08:12,563 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.969e+02 2.337e+02 2.871e+02 4.733e+02, threshold=4.673e+02, percent-clipped=1.0 +2023-03-21 02:08:28,718 INFO [train.py:901] (1/2) Epoch 22, batch 450, loss[loss=0.1628, simple_loss=0.237, pruned_loss=0.04427, over 7360.00 frames. ], tot_loss[loss=0.1487, simple_loss=0.2263, pruned_loss=0.03553, over 1292749.98 frames. ], batch size: 63, lr: 7.57e-03, grad_scale: 8.0 +2023-03-21 02:08:35,259 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 02:08:35,377 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9189, 2.1829, 3.1066, 2.7889, 2.8345, 2.6706, 2.4611, 2.9133], + device='cuda:1'), covar=tensor([0.1513, 0.1254, 0.0827, 0.1220, 0.1073, 0.1088, 0.2223, 0.1346], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0055, 0.0042, 0.0041, 0.0041, 0.0038, 0.0056, 0.0045], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:08:35,762 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 02:08:54,222 INFO [train.py:901] (1/2) Epoch 22, batch 500, loss[loss=0.1188, simple_loss=0.1956, pruned_loss=0.02105, over 7153.00 frames. ], tot_loss[loss=0.1498, simple_loss=0.2276, pruned_loss=0.03602, over 1327092.80 frames. ], batch size: 39, lr: 7.56e-03, grad_scale: 8.0 +2023-03-21 02:09:01,857 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59820.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:09:04,738 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.315e+02 1.908e+02 2.279e+02 2.698e+02 3.738e+02, threshold=4.558e+02, percent-clipped=0.0 +2023-03-21 02:09:08,393 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 02:09:09,055 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0141, 2.3979, 1.8979, 2.7113, 2.9643, 2.4430, 2.4279, 2.2138], + device='cuda:1'), covar=tensor([0.1739, 0.0803, 0.3033, 0.0542, 0.0151, 0.0113, 0.0222, 0.0258], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0234, 0.0266, 0.0262, 0.0165, 0.0165, 0.0194, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:09:09,876 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 02:09:10,409 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 02:09:12,490 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 02:09:17,229 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59848.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:09:18,135 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 02:09:20,649 INFO [train.py:901] (1/2) Epoch 22, batch 550, loss[loss=0.1542, simple_loss=0.2347, pruned_loss=0.0368, over 7267.00 frames. ], tot_loss[loss=0.1488, simple_loss=0.2268, pruned_loss=0.03541, over 1351437.09 frames. ], batch size: 86, lr: 7.56e-03, grad_scale: 8.0 +2023-03-21 02:09:22,288 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59858.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:09:27,307 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=59868.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:09:29,797 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 02:09:37,263 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 02:09:37,914 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59889.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:09:40,683 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 02:09:45,699 INFO [train.py:901] (1/2) Epoch 22, batch 600, loss[loss=0.1515, simple_loss=0.2329, pruned_loss=0.035, over 7294.00 frames. ], tot_loss[loss=0.1487, simple_loss=0.2267, pruned_loss=0.03537, over 1372730.08 frames. ], batch size: 68, lr: 7.56e-03, grad_scale: 8.0 +2023-03-21 02:09:48,346 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 02:09:48,496 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59909.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:09:56,960 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.308e+02 1.951e+02 2.184e+02 2.714e+02 3.801e+02, threshold=4.369e+02, percent-clipped=0.0 +2023-03-21 02:10:04,161 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4402, 3.6665, 3.4505, 3.6247, 3.3792, 3.5121, 3.8339, 3.8671], + device='cuda:1'), covar=tensor([0.0270, 0.0174, 0.0232, 0.0203, 0.0353, 0.0472, 0.0242, 0.0184], + device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0113, 0.0105, 0.0110, 0.0103, 0.0094, 0.0089, 0.0089], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:10:04,557 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 02:10:09,745 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59950.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:10:12,112 INFO [train.py:901] (1/2) Epoch 22, batch 650, loss[loss=0.1477, simple_loss=0.2265, pruned_loss=0.03448, over 7308.00 frames. ], tot_loss[loss=0.1485, simple_loss=0.2266, pruned_loss=0.03517, over 1387178.93 frames. ], batch size: 86, lr: 7.56e-03, grad_scale: 8.0 +2023-03-21 02:10:13,144 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 02:10:23,297 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6641, 3.1406, 3.6097, 3.7631, 3.6246, 3.7422, 3.6857, 3.5823], + device='cuda:1'), covar=tensor([0.0028, 0.0090, 0.0033, 0.0028, 0.0029, 0.0027, 0.0038, 0.0045], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0057, 0.0049, 0.0047, 0.0047, 0.0051, 0.0044, 0.0062], + device='cuda:1'), out_proj_covar=tensor([7.8643e-05, 1.3415e-04, 1.0650e-04, 9.4791e-05, 9.3630e-05, 1.0320e-04, + 9.8738e-05, 1.2982e-04], device='cuda:1') +2023-03-21 02:10:29,999 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-21 02:10:30,193 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 02:10:42,487 INFO [train.py:901] (1/2) Epoch 22, batch 700, loss[loss=0.1401, simple_loss=0.225, pruned_loss=0.02757, over 7212.00 frames. ], tot_loss[loss=0.1477, simple_loss=0.2261, pruned_loss=0.03467, over 1399732.13 frames. ], batch size: 50, lr: 7.55e-03, grad_scale: 8.0 +2023-03-21 02:10:44,476 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 02:10:48,611 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0883, 2.5569, 1.9279, 2.7809, 2.9651, 2.6741, 2.2949, 2.0996], + device='cuda:1'), covar=tensor([0.1699, 0.0819, 0.3139, 0.0605, 0.0155, 0.0100, 0.0225, 0.0252], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0233, 0.0264, 0.0261, 0.0164, 0.0163, 0.0194, 0.0208], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:10:52,395 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.325e+02 1.930e+02 2.266e+02 2.658e+02 4.058e+02, threshold=4.532e+02, percent-clipped=0.0 +2023-03-21 02:10:57,646 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0059, 2.4011, 1.8303, 2.6411, 3.0033, 2.6506, 2.2212, 2.2451], + device='cuda:1'), covar=tensor([0.1696, 0.0770, 0.3045, 0.0656, 0.0170, 0.0104, 0.0238, 0.0357], + device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0233, 0.0265, 0.0262, 0.0165, 0.0164, 0.0195, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:11:00,105 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60040.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:11:06,157 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60052.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:11:07,476 INFO [train.py:901] (1/2) Epoch 22, batch 750, loss[loss=0.1582, simple_loss=0.2279, pruned_loss=0.04419, over 7337.00 frames. ], tot_loss[loss=0.1482, simple_loss=0.2264, pruned_loss=0.03502, over 1411203.27 frames. ], batch size: 61, lr: 7.55e-03, grad_scale: 8.0 +2023-03-21 02:11:07,993 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 02:11:08,472 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 02:11:12,636 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60065.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:11:19,688 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60078.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:11:23,582 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 02:11:28,030 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.75 vs. limit=5.0 +2023-03-21 02:11:29,256 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 02:11:31,907 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60101.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 02:11:33,729 INFO [train.py:901] (1/2) Epoch 22, batch 800, loss[loss=0.142, simple_loss=0.2256, pruned_loss=0.02921, over 7271.00 frames. ], tot_loss[loss=0.1475, simple_loss=0.2255, pruned_loss=0.03476, over 1415350.71 frames. ], batch size: 70, lr: 7.55e-03, grad_scale: 8.0 +2023-03-21 02:11:34,444 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9943, 4.0137, 3.2814, 3.4505, 3.1012, 2.3331, 1.7147, 4.0267], + device='cuda:1'), covar=tensor([0.0043, 0.0034, 0.0109, 0.0066, 0.0118, 0.0478, 0.0588, 0.0047], + device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0076, 0.0096, 0.0082, 0.0108, 0.0121, 0.0119, 0.0088], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 02:11:35,325 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 02:11:36,348 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 02:11:37,876 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1407, 3.8202, 3.9158, 3.8548, 3.1306, 3.7436, 3.8208, 3.5767], + device='cuda:1'), covar=tensor([0.0210, 0.0210, 0.0160, 0.0211, 0.0734, 0.0206, 0.0316, 0.0242], + device='cuda:1'), in_proj_covar=tensor([0.0080, 0.0087, 0.0086, 0.0078, 0.0155, 0.0097, 0.0094, 0.0095], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:11:37,929 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60113.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 02:11:43,677 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 1.981e+02 2.306e+02 2.691e+02 7.119e+02, threshold=4.612e+02, percent-clipped=1.0 +2023-03-21 02:11:44,419 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60126.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:11:46,778 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 02:11:50,935 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60139.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:11:58,620 INFO [train.py:901] (1/2) Epoch 22, batch 850, loss[loss=0.1487, simple_loss=0.2285, pruned_loss=0.03441, over 7367.00 frames. ], tot_loss[loss=0.148, simple_loss=0.2258, pruned_loss=0.03509, over 1418925.61 frames. ], batch size: 73, lr: 7.54e-03, grad_scale: 8.0 +2023-03-21 02:12:00,267 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60158.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:12:05,794 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 02:12:05,803 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 02:12:11,868 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 02:12:14,887 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 02:12:21,408 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60198.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:12:24,625 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60204.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:12:25,053 INFO [train.py:901] (1/2) Epoch 22, batch 900, loss[loss=0.1613, simple_loss=0.2378, pruned_loss=0.04246, over 7255.00 frames. ], tot_loss[loss=0.1484, simple_loss=0.2265, pruned_loss=0.03517, over 1424659.65 frames. ], batch size: 64, lr: 7.54e-03, grad_scale: 8.0 +2023-03-21 02:12:25,605 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=60206.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:12:35,133 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.240e+02 1.857e+02 2.134e+02 2.609e+02 4.521e+02, threshold=4.269e+02, percent-clipped=0.0 +2023-03-21 02:12:45,228 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60245.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:12:50,760 INFO [train.py:901] (1/2) Epoch 22, batch 950, loss[loss=0.1239, simple_loss=0.1974, pruned_loss=0.0252, over 6969.00 frames. ], tot_loss[loss=0.1483, simple_loss=0.2264, pruned_loss=0.03509, over 1426577.75 frames. ], batch size: 35, lr: 7.54e-03, grad_scale: 8.0 +2023-03-21 02:12:52,707 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 02:12:53,454 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60259.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:13:00,600 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6331, 2.8432, 2.4826, 2.7456, 2.5998, 2.2768, 2.7506, 2.5670], + device='cuda:1'), covar=tensor([0.0768, 0.0554, 0.1200, 0.1444, 0.1047, 0.0849, 0.1058, 0.1468], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0048, 0.0055, 0.0049, 0.0047, 0.0049, 0.0050, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:13:16,262 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 02:13:16,718 INFO [train.py:901] (1/2) Epoch 22, batch 1000, loss[loss=0.1464, simple_loss=0.2272, pruned_loss=0.03278, over 7330.00 frames. ], tot_loss[loss=0.1477, simple_loss=0.2259, pruned_loss=0.03468, over 1429690.59 frames. ], batch size: 44, lr: 7.53e-03, grad_scale: 8.0 +2023-03-21 02:13:26,709 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.173e+02 1.855e+02 2.168e+02 2.543e+02 4.857e+02, threshold=4.336e+02, percent-clipped=1.0 +2023-03-21 02:13:36,619 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 02:13:42,176 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0848, 3.0116, 2.2481, 3.6544, 2.4903, 3.2769, 1.6853, 2.0744], + device='cuda:1'), covar=tensor([0.0367, 0.0742, 0.2267, 0.0591, 0.0428, 0.0544, 0.2907, 0.1814], + device='cuda:1'), in_proj_covar=tensor([0.0267, 0.0251, 0.0293, 0.0259, 0.0267, 0.0261, 0.0253, 0.0275], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], + device='cuda:1') +2023-03-21 02:13:42,924 INFO [train.py:901] (1/2) Epoch 22, batch 1050, loss[loss=0.1706, simple_loss=0.2456, pruned_loss=0.04779, over 7312.00 frames. ], tot_loss[loss=0.1478, simple_loss=0.2262, pruned_loss=0.03472, over 1434695.51 frames. ], batch size: 59, lr: 7.53e-03, grad_scale: 8.0 +2023-03-21 02:13:50,084 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7073, 3.1043, 2.4251, 3.7923, 3.7575, 3.6933, 3.4017, 3.3948], + device='cuda:1'), covar=tensor([0.1794, 0.0519, 0.3116, 0.0379, 0.0182, 0.0144, 0.0338, 0.0339], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0236, 0.0267, 0.0263, 0.0167, 0.0165, 0.0196, 0.0210], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:13:58,036 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 02:14:01,976 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 02:14:03,602 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60396.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 02:14:08,311 INFO [train.py:901] (1/2) Epoch 22, batch 1100, loss[loss=0.1374, simple_loss=0.2199, pruned_loss=0.02744, over 7291.00 frames. ], tot_loss[loss=0.1477, simple_loss=0.2261, pruned_loss=0.03468, over 1435505.63 frames. ], batch size: 86, lr: 7.53e-03, grad_scale: 8.0 +2023-03-21 02:14:09,881 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60408.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 02:14:15,564 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.09 vs. limit=5.0 +2023-03-21 02:14:17,044 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60421.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:14:19,040 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.201e+02 1.927e+02 2.269e+02 2.813e+02 4.355e+02, threshold=4.537e+02, percent-clipped=1.0 +2023-03-21 02:14:24,273 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60434.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:14:31,382 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 02:14:31,837 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 02:14:34,876 INFO [train.py:901] (1/2) Epoch 22, batch 1150, loss[loss=0.1519, simple_loss=0.2366, pruned_loss=0.03362, over 6668.00 frames. ], tot_loss[loss=0.1477, simple_loss=0.2262, pruned_loss=0.03461, over 1436313.65 frames. ], batch size: 106, lr: 7.52e-03, grad_scale: 8.0 +2023-03-21 02:14:43,412 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 02:14:44,365 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 02:14:51,009 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1071, 2.4947, 2.1056, 3.7520, 1.6293, 3.5576, 1.4045, 2.6905], + device='cuda:1'), covar=tensor([0.0117, 0.0971, 0.1604, 0.0111, 0.3747, 0.0183, 0.1195, 0.0389], + device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0261, 0.0278, 0.0184, 0.0266, 0.0192, 0.0253, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:14:59,442 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60504.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:14:59,845 INFO [train.py:901] (1/2) Epoch 22, batch 1200, loss[loss=0.1741, simple_loss=0.2516, pruned_loss=0.04834, over 6719.00 frames. ], tot_loss[loss=0.1476, simple_loss=0.2259, pruned_loss=0.0346, over 1435958.60 frames. ], batch size: 106, lr: 7.52e-03, grad_scale: 8.0 +2023-03-21 02:15:11,230 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.263e+02 2.000e+02 2.319e+02 2.999e+02 4.801e+02, threshold=4.638e+02, percent-clipped=2.0 +2023-03-21 02:15:17,702 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 02:15:21,398 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60545.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:15:24,833 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=60552.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:15:25,886 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60554.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:15:26,302 INFO [train.py:901] (1/2) Epoch 22, batch 1250, loss[loss=0.1463, simple_loss=0.2247, pruned_loss=0.03397, over 7297.00 frames. ], tot_loss[loss=0.1471, simple_loss=0.2253, pruned_loss=0.0344, over 1436606.86 frames. ], batch size: 70, lr: 7.52e-03, grad_scale: 8.0 +2023-03-21 02:15:28,962 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0848, 2.5591, 3.1397, 2.9569, 3.0031, 2.8546, 2.6354, 3.1594], + device='cuda:1'), covar=tensor([0.1739, 0.0652, 0.1168, 0.1863, 0.1003, 0.1101, 0.2109, 0.1198], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0056, 0.0043, 0.0043, 0.0043, 0.0040, 0.0058, 0.0045], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:15:40,274 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 02:15:43,838 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 02:15:45,409 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 02:15:45,965 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=60593.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:15:52,405 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6796, 3.3926, 3.6011, 3.6201, 2.9524, 3.2517, 3.6537, 3.2735], + device='cuda:1'), covar=tensor([0.0385, 0.0327, 0.0236, 0.0301, 0.1011, 0.0341, 0.0337, 0.0345], + device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0087, 0.0084, 0.0077, 0.0153, 0.0096, 0.0092, 0.0093], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:15:52,802 INFO [train.py:901] (1/2) Epoch 22, batch 1300, loss[loss=0.1641, simple_loss=0.2365, pruned_loss=0.04583, over 7331.00 frames. ], tot_loss[loss=0.1477, simple_loss=0.2264, pruned_loss=0.03453, over 1439421.96 frames. ], batch size: 61, lr: 7.51e-03, grad_scale: 8.0 +2023-03-21 02:16:02,811 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.367e+02 2.027e+02 2.390e+02 2.761e+02 3.911e+02, threshold=4.781e+02, percent-clipped=0.0 +2023-03-21 02:16:08,605 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4061, 0.8706, 1.6174, 1.7514, 1.4899, 1.8280, 1.2347, 1.7136], + device='cuda:1'), covar=tensor([0.1974, 0.4482, 0.0833, 0.1089, 0.1908, 0.1505, 0.2167, 0.2163], + device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0064, 0.0050, 0.0045, 0.0048, 0.0051, 0.0074, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:16:09,453 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 02:16:11,495 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 02:16:14,826 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 +2023-03-21 02:16:15,038 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 02:16:18,039 INFO [train.py:901] (1/2) Epoch 22, batch 1350, loss[loss=0.1504, simple_loss=0.2298, pruned_loss=0.03556, over 7297.00 frames. ], tot_loss[loss=0.1477, simple_loss=0.2262, pruned_loss=0.03457, over 1438897.87 frames. ], batch size: 86, lr: 7.51e-03, grad_scale: 8.0 +2023-03-21 02:16:22,095 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0027, 3.6437, 3.8020, 3.7188, 3.6044, 3.5119, 3.9219, 3.4661], + device='cuda:1'), covar=tensor([0.0130, 0.0174, 0.0114, 0.0168, 0.0434, 0.0148, 0.0140, 0.0171], + device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0088, 0.0085, 0.0077, 0.0155, 0.0096, 0.0093, 0.0094], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:16:25,985 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 02:16:28,298 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 +2023-03-21 02:16:40,040 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60696.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:16:44,541 INFO [train.py:901] (1/2) Epoch 22, batch 1400, loss[loss=0.1284, simple_loss=0.1963, pruned_loss=0.03023, over 7013.00 frames. ], tot_loss[loss=0.1474, simple_loss=0.2258, pruned_loss=0.0345, over 1440002.65 frames. ], batch size: 35, lr: 7.51e-03, grad_scale: 8.0 +2023-03-21 02:16:46,155 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60708.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:16:51,232 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60718.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:16:52,618 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60721.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:16:54,500 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.546e+02 1.876e+02 2.210e+02 2.686e+02 7.585e+02, threshold=4.419e+02, percent-clipped=2.0 +2023-03-21 02:16:59,041 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60734.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:16:59,443 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 02:17:04,100 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=60744.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:17:09,483 INFO [train.py:901] (1/2) Epoch 22, batch 1450, loss[loss=0.1615, simple_loss=0.2419, pruned_loss=0.04049, over 7219.00 frames. ], tot_loss[loss=0.1474, simple_loss=0.2257, pruned_loss=0.03456, over 1439459.88 frames. ], batch size: 93, lr: 7.51e-03, grad_scale: 8.0 +2023-03-21 02:17:10,050 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=60756.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:17:13,678 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3949, 2.5120, 2.1557, 2.4897, 2.3372, 1.9543, 2.5229, 2.3664], + device='cuda:1'), covar=tensor([0.0833, 0.0550, 0.1429, 0.1008, 0.0932, 0.1141, 0.0726, 0.1037], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0048, 0.0055, 0.0048, 0.0046, 0.0049, 0.0049, 0.0045], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:17:17,102 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=60769.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:17:22,232 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60779.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:17:24,153 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 02:17:24,197 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=60782.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:17:25,299 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8050, 2.9816, 2.5366, 3.0267, 2.8108, 2.4478, 2.9455, 2.8521], + device='cuda:1'), covar=tensor([0.0892, 0.0622, 0.1251, 0.0729, 0.1192, 0.0903, 0.1132, 0.1113], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0048, 0.0055, 0.0048, 0.0046, 0.0049, 0.0049, 0.0045], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:17:35,351 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 +2023-03-21 02:17:36,006 INFO [train.py:901] (1/2) Epoch 22, batch 1500, loss[loss=0.1343, simple_loss=0.2205, pruned_loss=0.02398, over 7334.00 frames. ], tot_loss[loss=0.1476, simple_loss=0.2261, pruned_loss=0.0346, over 1442292.37 frames. ], batch size: 44, lr: 7.50e-03, grad_scale: 8.0 +2023-03-21 02:17:39,961 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 02:17:41,140 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3736, 0.9020, 1.6144, 1.8118, 1.5237, 1.8791, 1.3299, 1.8014], + device='cuda:1'), covar=tensor([0.1878, 0.2401, 0.0728, 0.1514, 0.2115, 0.0875, 0.1652, 0.2617], + device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0062, 0.0050, 0.0045, 0.0047, 0.0050, 0.0073, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:17:45,916 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.359e+02 2.001e+02 2.411e+02 2.741e+02 4.472e+02, threshold=4.822e+02, percent-clipped=1.0 +2023-03-21 02:18:00,607 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60854.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:18:01,625 INFO [train.py:901] (1/2) Epoch 22, batch 1550, loss[loss=0.1471, simple_loss=0.2255, pruned_loss=0.03435, over 7284.00 frames. ], tot_loss[loss=0.1477, simple_loss=0.2256, pruned_loss=0.03488, over 1440492.41 frames. ], batch size: 66, lr: 7.50e-03, grad_scale: 8.0 +2023-03-21 02:18:04,658 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 02:18:25,809 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=60902.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:18:27,246 INFO [train.py:901] (1/2) Epoch 22, batch 1600, loss[loss=0.1561, simple_loss=0.2339, pruned_loss=0.0391, over 7362.00 frames. ], tot_loss[loss=0.1477, simple_loss=0.2259, pruned_loss=0.03474, over 1443026.62 frames. ], batch size: 73, lr: 7.50e-03, grad_scale: 8.0 +2023-03-21 02:18:34,810 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 02:18:35,773 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 02:18:37,319 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.373e+02 1.951e+02 2.358e+02 2.775e+02 4.018e+02, threshold=4.716e+02, percent-clipped=0.0 +2023-03-21 02:18:38,360 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 02:18:49,095 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 02:18:53,670 INFO [train.py:901] (1/2) Epoch 22, batch 1650, loss[loss=0.1301, simple_loss=0.2004, pruned_loss=0.02993, over 7228.00 frames. ], tot_loss[loss=0.1478, simple_loss=0.2259, pruned_loss=0.03482, over 1441639.20 frames. ], batch size: 45, lr: 7.49e-03, grad_scale: 8.0 +2023-03-21 02:18:53,692 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 02:18:56,182 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-21 02:18:57,102 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-21 02:19:01,423 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3753, 4.8726, 4.9547, 4.9157, 4.7587, 4.3382, 5.0014, 4.7541], + device='cuda:1'), covar=tensor([0.0396, 0.0393, 0.0346, 0.0365, 0.0331, 0.0355, 0.0275, 0.0457], + device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0226, 0.0171, 0.0170, 0.0136, 0.0203, 0.0179, 0.0135], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 02:19:02,349 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 02:19:19,364 INFO [train.py:901] (1/2) Epoch 22, batch 1700, loss[loss=0.1498, simple_loss=0.2294, pruned_loss=0.03509, over 7336.00 frames. ], tot_loss[loss=0.1478, simple_loss=0.2258, pruned_loss=0.03486, over 1443388.73 frames. ], batch size: 54, lr: 7.49e-03, grad_scale: 16.0 +2023-03-21 02:19:19,401 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 02:19:23,971 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 02:19:30,089 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.186e+02 1.914e+02 2.304e+02 2.748e+02 4.764e+02, threshold=4.609e+02, percent-clipped=1.0 +2023-03-21 02:19:35,151 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 02:19:44,970 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2423, 3.4020, 3.5332, 3.1894, 3.7407, 3.2309, 2.3870, 3.2993], + device='cuda:1'), covar=tensor([0.2161, 0.0938, 0.1044, 0.1271, 0.0695, 0.2111, 0.3225, 0.2215], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0055, 0.0042, 0.0041, 0.0042, 0.0039, 0.0057, 0.0045], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:19:45,763 INFO [train.py:901] (1/2) Epoch 22, batch 1750, loss[loss=0.1586, simple_loss=0.2408, pruned_loss=0.03818, over 7278.00 frames. ], tot_loss[loss=0.1469, simple_loss=0.2252, pruned_loss=0.0343, over 1442721.32 frames. ], batch size: 64, lr: 7.49e-03, grad_scale: 16.0 +2023-03-21 02:19:48,896 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61061.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:19:55,502 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61074.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:20:00,892 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 02:20:01,918 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 02:20:10,965 INFO [train.py:901] (1/2) Epoch 22, batch 1800, loss[loss=0.1398, simple_loss=0.2224, pruned_loss=0.02859, over 7301.00 frames. ], tot_loss[loss=0.1468, simple_loss=0.225, pruned_loss=0.03428, over 1441001.07 frames. ], batch size: 80, lr: 7.48e-03, grad_scale: 16.0 +2023-03-21 02:20:20,040 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61122.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:20:21,379 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.437e+02 1.953e+02 2.262e+02 2.819e+02 5.491e+02, threshold=4.524e+02, percent-clipped=2.0 +2023-03-21 02:20:22,948 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 02:20:28,621 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3460, 1.6246, 1.3387, 1.4478, 1.5801, 1.4359, 1.4843, 1.2274], + device='cuda:1'), covar=tensor([0.0250, 0.0132, 0.0260, 0.0177, 0.0129, 0.0131, 0.0138, 0.0129], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0025, 0.0025, 0.0028, 0.0024, 0.0025, 0.0027, 0.0034], + device='cuda:1'), out_proj_covar=tensor([3.1996e-05, 2.8749e-05, 2.9356e-05, 3.0759e-05, 2.7687e-05, 2.8191e-05, + 3.1150e-05, 3.8421e-05], device='cuda:1') +2023-03-21 02:20:37,058 INFO [train.py:901] (1/2) Epoch 22, batch 1850, loss[loss=0.1507, simple_loss=0.2308, pruned_loss=0.0353, over 7302.00 frames. ], tot_loss[loss=0.1473, simple_loss=0.2255, pruned_loss=0.03451, over 1442894.69 frames. ], batch size: 86, lr: 7.48e-03, grad_scale: 8.0 +2023-03-21 02:20:37,064 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 02:20:47,560 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 02:21:03,042 INFO [train.py:901] (1/2) Epoch 22, batch 1900, loss[loss=0.1241, simple_loss=0.1812, pruned_loss=0.0335, over 5772.00 frames. ], tot_loss[loss=0.1472, simple_loss=0.2256, pruned_loss=0.03442, over 1441966.47 frames. ], batch size: 25, lr: 7.48e-03, grad_scale: 8.0 +2023-03-21 02:21:04,552 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 02:21:09,736 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7188, 3.3983, 3.6840, 3.5174, 3.2046, 2.9297, 3.8845, 2.8475], + device='cuda:1'), covar=tensor([0.0434, 0.0376, 0.0363, 0.0430, 0.0579, 0.0801, 0.0444, 0.1268], + device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0330, 0.0266, 0.0354, 0.0305, 0.0301, 0.0341, 0.0286], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:21:14,168 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 1.892e+02 2.344e+02 2.726e+02 6.407e+02, threshold=4.688e+02, percent-clipped=1.0 +2023-03-21 02:21:24,300 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6749, 2.7073, 2.1312, 3.4940, 3.5112, 3.1490, 3.1190, 3.0092], + device='cuda:1'), covar=tensor([0.1726, 0.0585, 0.3196, 0.0489, 0.0142, 0.0107, 0.0280, 0.0317], + device='cuda:1'), in_proj_covar=tensor([0.0265, 0.0233, 0.0261, 0.0261, 0.0165, 0.0166, 0.0195, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:21:28,656 INFO [train.py:901] (1/2) Epoch 22, batch 1950, loss[loss=0.1373, simple_loss=0.2209, pruned_loss=0.02681, over 7257.00 frames. ], tot_loss[loss=0.1481, simple_loss=0.2264, pruned_loss=0.0349, over 1441606.03 frames. ], batch size: 55, lr: 7.47e-03, grad_scale: 8.0 +2023-03-21 02:21:29,182 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 02:21:39,844 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6682, 1.8137, 1.9028, 1.7702, 1.8825, 1.7947, 1.4746, 1.5854], + device='cuda:1'), covar=tensor([0.0536, 0.0446, 0.0200, 0.0210, 0.0316, 0.0481, 0.0388, 0.0292], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0027, 0.0026, 0.0026, 0.0028, 0.0027, 0.0029, 0.0029], + device='cuda:1'), out_proj_covar=tensor([7.2378e-05, 7.1320e-05, 6.6943e-05, 6.6751e-05, 7.0376e-05, 6.8261e-05, + 7.2489e-05, 7.5027e-05], device='cuda:1') +2023-03-21 02:21:40,227 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 02:21:40,325 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3716, 3.9580, 4.0609, 4.0929, 3.9555, 3.9203, 4.3243, 3.8104], + device='cuda:1'), covar=tensor([0.0108, 0.0139, 0.0122, 0.0116, 0.0433, 0.0128, 0.0114, 0.0162], + device='cuda:1'), in_proj_covar=tensor([0.0080, 0.0089, 0.0087, 0.0077, 0.0156, 0.0097, 0.0093, 0.0095], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:21:45,336 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 02:21:45,836 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 02:21:54,377 INFO [train.py:901] (1/2) Epoch 22, batch 2000, loss[loss=0.1508, simple_loss=0.2301, pruned_loss=0.03578, over 7347.00 frames. ], tot_loss[loss=0.1486, simple_loss=0.2268, pruned_loss=0.03517, over 1440591.81 frames. ], batch size: 61, lr: 7.47e-03, grad_scale: 8.0 +2023-03-21 02:22:01,178 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3195, 4.8638, 4.8688, 4.8299, 4.7359, 4.4289, 4.9515, 4.8014], + device='cuda:1'), covar=tensor([0.0431, 0.0353, 0.0351, 0.0420, 0.0270, 0.0344, 0.0275, 0.0356], + device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0224, 0.0173, 0.0172, 0.0136, 0.0205, 0.0181, 0.0135], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 02:22:02,590 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 02:22:03,743 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7061, 3.8537, 3.6183, 3.8196, 3.5129, 3.7513, 4.0699, 4.0537], + device='cuda:1'), covar=tensor([0.0224, 0.0184, 0.0253, 0.0200, 0.0339, 0.0381, 0.0255, 0.0207], + device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0114, 0.0106, 0.0111, 0.0105, 0.0095, 0.0092, 0.0090], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:22:03,836 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9485, 2.6065, 2.9795, 2.7238, 2.6216, 2.4254, 2.9986, 2.4244], + device='cuda:1'), covar=tensor([0.0428, 0.0424, 0.0456, 0.0401, 0.0462, 0.0677, 0.0406, 0.1353], + device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0328, 0.0264, 0.0352, 0.0302, 0.0301, 0.0340, 0.0285], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:22:05,613 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 1.924e+02 2.207e+02 2.701e+02 5.428e+02, threshold=4.414e+02, percent-clipped=1.0 +2023-03-21 02:22:12,584 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 02:22:20,029 INFO [train.py:901] (1/2) Epoch 22, batch 2050, loss[loss=0.1604, simple_loss=0.2407, pruned_loss=0.04004, over 7124.00 frames. ], tot_loss[loss=0.1489, simple_loss=0.2272, pruned_loss=0.03527, over 1442614.63 frames. ], batch size: 98, lr: 7.47e-03, grad_scale: 8.0 +2023-03-21 02:22:21,064 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 02:22:28,733 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9655, 2.7902, 3.1826, 2.9739, 3.2605, 2.8981, 2.4462, 3.2733], + device='cuda:1'), covar=tensor([0.1954, 0.0640, 0.1255, 0.1383, 0.0788, 0.1117, 0.2155, 0.1463], + device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0056, 0.0043, 0.0041, 0.0042, 0.0039, 0.0058, 0.0045], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:22:30,214 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61374.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:22:46,756 INFO [train.py:901] (1/2) Epoch 22, batch 2100, loss[loss=0.1065, simple_loss=0.1844, pruned_loss=0.01431, over 7001.00 frames. ], tot_loss[loss=0.1484, simple_loss=0.2271, pruned_loss=0.03485, over 1443876.37 frames. ], batch size: 35, lr: 7.47e-03, grad_scale: 8.0 +2023-03-21 02:22:52,725 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61417.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:22:54,148 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 02:22:55,228 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=61422.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:22:57,129 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.451e+02 2.046e+02 2.349e+02 2.806e+02 4.447e+02, threshold=4.697e+02, percent-clipped=1.0 +2023-03-21 02:22:57,676 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 02:23:11,617 INFO [train.py:901] (1/2) Epoch 22, batch 2150, loss[loss=0.1521, simple_loss=0.2261, pruned_loss=0.03903, over 7251.00 frames. ], tot_loss[loss=0.1486, simple_loss=0.227, pruned_loss=0.0351, over 1440319.98 frames. ], batch size: 55, lr: 7.46e-03, grad_scale: 8.0 +2023-03-21 02:23:17,949 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0406, 3.9503, 3.1582, 3.5211, 3.1482, 2.2377, 1.9186, 4.0880], + device='cuda:1'), covar=tensor([0.0049, 0.0044, 0.0125, 0.0071, 0.0121, 0.0472, 0.0537, 0.0049], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0076, 0.0098, 0.0085, 0.0110, 0.0121, 0.0122, 0.0090], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 02:23:38,225 INFO [train.py:901] (1/2) Epoch 22, batch 2200, loss[loss=0.1663, simple_loss=0.2456, pruned_loss=0.04346, over 7362.00 frames. ], tot_loss[loss=0.1486, simple_loss=0.2272, pruned_loss=0.03505, over 1439983.71 frames. ], batch size: 51, lr: 7.46e-03, grad_scale: 8.0 +2023-03-21 02:23:42,228 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 02:23:48,792 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.260e+02 1.877e+02 2.439e+02 2.820e+02 5.772e+02, threshold=4.878e+02, percent-clipped=2.0 +2023-03-21 02:24:03,356 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 +2023-03-21 02:24:03,963 INFO [train.py:901] (1/2) Epoch 22, batch 2250, loss[loss=0.1434, simple_loss=0.2189, pruned_loss=0.03396, over 7329.00 frames. ], tot_loss[loss=0.1473, simple_loss=0.2257, pruned_loss=0.03448, over 1438962.77 frames. ], batch size: 44, lr: 7.46e-03, grad_scale: 8.0 +2023-03-21 02:24:17,266 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 02:24:17,279 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 02:24:29,202 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 02:24:30,181 INFO [train.py:901] (1/2) Epoch 22, batch 2300, loss[loss=0.1522, simple_loss=0.2313, pruned_loss=0.03653, over 7361.00 frames. ], tot_loss[loss=0.1479, simple_loss=0.2263, pruned_loss=0.03471, over 1442492.68 frames. ], batch size: 73, lr: 7.45e-03, grad_scale: 8.0 +2023-03-21 02:24:40,705 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.277e+02 1.952e+02 2.374e+02 2.759e+02 5.347e+02, threshold=4.748e+02, percent-clipped=1.0 +2023-03-21 02:24:56,610 INFO [train.py:901] (1/2) Epoch 22, batch 2350, loss[loss=0.1418, simple_loss=0.2288, pruned_loss=0.0274, over 7210.00 frames. ], tot_loss[loss=0.1478, simple_loss=0.2261, pruned_loss=0.03469, over 1440973.49 frames. ], batch size: 93, lr: 7.45e-03, grad_scale: 8.0 +2023-03-21 02:25:16,086 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 02:25:21,583 INFO [train.py:901] (1/2) Epoch 22, batch 2400, loss[loss=0.1389, simple_loss=0.2205, pruned_loss=0.02864, over 7365.00 frames. ], tot_loss[loss=0.1472, simple_loss=0.2258, pruned_loss=0.03434, over 1442441.37 frames. ], batch size: 51, lr: 7.45e-03, grad_scale: 8.0 +2023-03-21 02:25:21,618 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 02:25:28,409 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61717.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:25:32,927 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.427e+02 1.971e+02 2.295e+02 2.794e+02 4.281e+02, threshold=4.590e+02, percent-clipped=0.0 +2023-03-21 02:25:33,446 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 02:25:35,993 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 02:25:36,096 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61732.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:25:42,286 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61743.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:25:48,401 INFO [train.py:901] (1/2) Epoch 22, batch 2450, loss[loss=0.1312, simple_loss=0.218, pruned_loss=0.02218, over 7314.00 frames. ], tot_loss[loss=0.1469, simple_loss=0.2252, pruned_loss=0.03428, over 1440157.96 frames. ], batch size: 59, lr: 7.44e-03, grad_scale: 8.0 +2023-03-21 02:25:53,486 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=61765.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:25:53,612 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8682, 2.8525, 2.1617, 3.5825, 2.4467, 2.9957, 1.5391, 2.1282], + device='cuda:1'), covar=tensor([0.0350, 0.0617, 0.2262, 0.0550, 0.0580, 0.0411, 0.2649, 0.1608], + device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0248, 0.0289, 0.0258, 0.0264, 0.0259, 0.0250, 0.0271], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:26:03,074 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 02:26:07,671 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61793.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:26:11,531 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1388, 2.1065, 2.1684, 3.3441, 1.5842, 3.3025, 1.3596, 2.7012], + device='cuda:1'), covar=tensor([0.0111, 0.1210, 0.1652, 0.0116, 0.3935, 0.0129, 0.1178, 0.0311], + device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0263, 0.0280, 0.0185, 0.0268, 0.0194, 0.0255, 0.0229], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:26:14,146 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61804.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:26:14,507 INFO [train.py:901] (1/2) Epoch 22, batch 2500, loss[loss=0.1648, simple_loss=0.2432, pruned_loss=0.04315, over 7297.00 frames. ], tot_loss[loss=0.1471, simple_loss=0.2254, pruned_loss=0.03441, over 1439920.31 frames. ], batch size: 86, lr: 7.44e-03, grad_scale: 8.0 +2023-03-21 02:26:25,716 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.431e+02 1.926e+02 2.287e+02 2.628e+02 4.851e+02, threshold=4.574e+02, percent-clipped=1.0 +2023-03-21 02:26:29,180 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 02:26:40,367 INFO [train.py:901] (1/2) Epoch 22, batch 2550, loss[loss=0.1321, simple_loss=0.2127, pruned_loss=0.02578, over 7162.00 frames. ], tot_loss[loss=0.1469, simple_loss=0.2255, pruned_loss=0.03414, over 1441954.94 frames. ], batch size: 41, lr: 7.44e-03, grad_scale: 8.0 +2023-03-21 02:26:44,744 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 02:26:59,806 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.11 vs. limit=2.0 +2023-03-21 02:27:05,956 INFO [train.py:901] (1/2) Epoch 22, batch 2600, loss[loss=0.1203, simple_loss=0.1842, pruned_loss=0.02822, over 5865.00 frames. ], tot_loss[loss=0.1473, simple_loss=0.2257, pruned_loss=0.03438, over 1440169.33 frames. ], batch size: 25, lr: 7.44e-03, grad_scale: 8.0 +2023-03-21 02:27:07,104 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9920, 2.1280, 2.0875, 3.2127, 1.5393, 3.1867, 1.3264, 2.5547], + device='cuda:1'), covar=tensor([0.0135, 0.1058, 0.1651, 0.0151, 0.3822, 0.0149, 0.1259, 0.0269], + device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0262, 0.0280, 0.0186, 0.0265, 0.0194, 0.0254, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:27:10,478 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61914.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:27:14,463 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9350, 3.1355, 2.2615, 3.8271, 2.5936, 3.3251, 1.6914, 2.1566], + device='cuda:1'), covar=tensor([0.0266, 0.0599, 0.2000, 0.0544, 0.0352, 0.0473, 0.2695, 0.1531], + device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0248, 0.0289, 0.0258, 0.0265, 0.0257, 0.0250, 0.0271], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:27:16,188 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.332e+02 1.945e+02 2.250e+02 2.728e+02 4.156e+02, threshold=4.499e+02, percent-clipped=0.0 +2023-03-21 02:27:31,130 INFO [train.py:901] (1/2) Epoch 22, batch 2650, loss[loss=0.1395, simple_loss=0.2324, pruned_loss=0.02331, over 7255.00 frames. ], tot_loss[loss=0.147, simple_loss=0.2254, pruned_loss=0.03427, over 1442428.38 frames. ], batch size: 89, lr: 7.43e-03, grad_scale: 8.0 +2023-03-21 02:27:41,116 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61975.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:27:52,203 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 +2023-03-21 02:27:56,100 INFO [train.py:901] (1/2) Epoch 22, batch 2700, loss[loss=0.1466, simple_loss=0.2256, pruned_loss=0.03382, over 7262.00 frames. ], tot_loss[loss=0.1472, simple_loss=0.2255, pruned_loss=0.03452, over 1440738.81 frames. ], batch size: 47, lr: 7.43e-03, grad_scale: 8.0 +2023-03-21 02:27:56,835 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 +2023-03-21 02:28:06,226 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.177e+02 2.000e+02 2.344e+02 2.679e+02 5.407e+02, threshold=4.688e+02, percent-clipped=1.0 +2023-03-21 02:28:20,670 INFO [train.py:901] (1/2) Epoch 22, batch 2750, loss[loss=0.1466, simple_loss=0.2284, pruned_loss=0.03242, over 7308.00 frames. ], tot_loss[loss=0.1471, simple_loss=0.2251, pruned_loss=0.0345, over 1439335.98 frames. ], batch size: 59, lr: 7.43e-03, grad_scale: 8.0 +2023-03-21 02:28:36,983 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62088.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:28:38,686 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.48 vs. limit=5.0 +2023-03-21 02:28:42,356 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62099.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:28:43,969 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8580, 2.1762, 1.6589, 2.5460, 2.7012, 2.5824, 2.2587, 2.4863], + device='cuda:1'), covar=tensor([0.1720, 0.0789, 0.3051, 0.0626, 0.0192, 0.0193, 0.0314, 0.0277], + device='cuda:1'), in_proj_covar=tensor([0.0264, 0.0231, 0.0263, 0.0262, 0.0168, 0.0169, 0.0198, 0.0209], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:28:44,887 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3378, 0.9872, 1.5153, 1.7417, 1.5315, 1.7892, 1.1846, 1.6453], + device='cuda:1'), covar=tensor([0.2559, 0.3050, 0.2077, 0.2760, 0.2246, 0.1734, 0.1563, 0.2376], + device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0063, 0.0050, 0.0046, 0.0048, 0.0049, 0.0073, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:28:45,247 INFO [train.py:901] (1/2) Epoch 22, batch 2800, loss[loss=0.1652, simple_loss=0.2474, pruned_loss=0.04152, over 7317.00 frames. ], tot_loss[loss=0.1475, simple_loss=0.2258, pruned_loss=0.03461, over 1440436.04 frames. ], batch size: 83, lr: 7.42e-03, grad_scale: 8.0 +2023-03-21 02:28:53,556 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62122.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:28:55,322 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 1.965e+02 2.306e+02 2.892e+02 4.105e+02, threshold=4.613e+02, percent-clipped=0.0 +2023-03-21 02:29:10,722 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 02:29:10,777 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 02:29:10,801 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 02:29:10,818 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 02:29:11,288 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 02:29:11,303 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 02:29:11,314 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 02:29:11,415 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 02:29:11,488 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 02:29:11,503 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 02:29:11,520 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 02:29:11,680 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 02:29:11,687 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 02:29:11,694 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 02:29:11,832 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 02:29:11,835 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 02:29:12,244 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 02:29:20,238 INFO [train.py:901] (1/2) Epoch 23, batch 0, loss[loss=0.1508, simple_loss=0.2298, pruned_loss=0.03585, over 7296.00 frames. ], tot_loss[loss=0.1508, simple_loss=0.2298, pruned_loss=0.03585, over 7296.00 frames. ], batch size: 86, lr: 7.26e-03, grad_scale: 8.0 +2023-03-21 02:29:20,239 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 02:29:36,332 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6156, 2.8385, 2.7221, 2.9764, 2.7877, 2.5955, 3.0273, 2.8258], + device='cuda:1'), covar=tensor([0.1084, 0.1008, 0.0678, 0.0920, 0.1059, 0.0657, 0.0573, 0.0575], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0050, 0.0057, 0.0051, 0.0049, 0.0051, 0.0051, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:29:45,949 INFO [train.py:935] (1/2) Epoch 23, validation: loss=0.1647, simple_loss=0.2529, pruned_loss=0.03826, over 1622729.00 frames. +2023-03-21 02:29:45,950 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 02:29:50,901 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 +2023-03-21 02:29:53,079 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 02:29:54,536 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-21 02:30:04,951 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 02:30:11,901 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 02:30:12,384 INFO [train.py:901] (1/2) Epoch 23, batch 50, loss[loss=0.1288, simple_loss=0.216, pruned_loss=0.02079, over 7319.00 frames. ], tot_loss[loss=0.1498, simple_loss=0.2292, pruned_loss=0.03515, over 326013.00 frames. ], batch size: 59, lr: 7.26e-03, grad_scale: 8.0 +2023-03-21 02:30:13,866 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 02:30:14,500 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62183.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:30:16,285 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 02:30:21,398 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0179, 2.7124, 3.0914, 3.0686, 3.3842, 2.8096, 2.6679, 3.3440], + device='cuda:1'), covar=tensor([0.1653, 0.0772, 0.1399, 0.1463, 0.0735, 0.1113, 0.2211, 0.0993], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0057, 0.0044, 0.0044, 0.0044, 0.0042, 0.0060, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:30:23,323 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8467, 2.6299, 2.8862, 2.9107, 2.6662, 2.4190, 2.8576, 2.3000], + device='cuda:1'), covar=tensor([0.0455, 0.0434, 0.0410, 0.0574, 0.0574, 0.0792, 0.0502, 0.1531], + device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0332, 0.0265, 0.0357, 0.0306, 0.0303, 0.0345, 0.0289], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:30:24,279 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9415, 2.8409, 2.4655, 3.7305, 1.7013, 3.5710, 1.8229, 2.6258], + device='cuda:1'), covar=tensor([0.0116, 0.0844, 0.1478, 0.0120, 0.3483, 0.0176, 0.1070, 0.0263], + device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0263, 0.0280, 0.0186, 0.0267, 0.0195, 0.0253, 0.0229], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:30:26,710 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1099, 4.1335, 3.4107, 3.6324, 3.0927, 2.2209, 1.8399, 4.1608], + device='cuda:1'), covar=tensor([0.0038, 0.0035, 0.0081, 0.0064, 0.0122, 0.0492, 0.0581, 0.0047], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0077, 0.0099, 0.0085, 0.0112, 0.0123, 0.0124, 0.0091], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 02:30:32,090 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 02:30:33,091 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 02:30:34,845 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.50 vs. limit=5.0 +2023-03-21 02:30:36,062 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.420e+02 2.038e+02 2.389e+02 2.944e+02 6.721e+02, threshold=4.778e+02, percent-clipped=2.0 +2023-03-21 02:30:37,549 INFO [train.py:901] (1/2) Epoch 23, batch 100, loss[loss=0.1499, simple_loss=0.2281, pruned_loss=0.03586, over 7277.00 frames. ], tot_loss[loss=0.1491, simple_loss=0.2282, pruned_loss=0.03504, over 574193.00 frames. ], batch size: 89, lr: 7.26e-03, grad_scale: 8.0 +2023-03-21 02:30:43,153 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2814, 4.7384, 4.7675, 4.6992, 4.6496, 4.3096, 4.8409, 4.6358], + device='cuda:1'), covar=tensor([0.0406, 0.0386, 0.0411, 0.0506, 0.0278, 0.0358, 0.0314, 0.0440], + device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0226, 0.0177, 0.0174, 0.0137, 0.0208, 0.0181, 0.0135], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 02:30:53,368 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2706, 1.4677, 1.3034, 1.4457, 1.5124, 1.4120, 1.4922, 1.0583], + device='cuda:1'), covar=tensor([0.0177, 0.0106, 0.0162, 0.0114, 0.0074, 0.0122, 0.0099, 0.0122], + device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0025, 0.0025, 0.0027, 0.0024, 0.0025, 0.0027, 0.0033], + device='cuda:1'), out_proj_covar=tensor([3.1059e-05, 2.8791e-05, 2.9244e-05, 2.9722e-05, 2.7572e-05, 2.7771e-05, + 3.0530e-05, 3.7735e-05], device='cuda:1') +2023-03-21 02:30:59,319 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62270.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:31:00,859 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62273.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:31:03,795 INFO [train.py:901] (1/2) Epoch 23, batch 150, loss[loss=0.1332, simple_loss=0.2195, pruned_loss=0.02339, over 7331.00 frames. ], tot_loss[loss=0.1471, simple_loss=0.2257, pruned_loss=0.03421, over 766961.86 frames. ], batch size: 54, lr: 7.26e-03, grad_scale: 8.0 +2023-03-21 02:31:28,324 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.231e+02 1.982e+02 2.401e+02 2.805e+02 5.339e+02, threshold=4.801e+02, percent-clipped=1.0 +2023-03-21 02:31:29,900 INFO [train.py:901] (1/2) Epoch 23, batch 200, loss[loss=0.1122, simple_loss=0.1723, pruned_loss=0.02608, over 5937.00 frames. ], tot_loss[loss=0.1468, simple_loss=0.2257, pruned_loss=0.03401, over 916009.42 frames. ], batch size: 25, lr: 7.25e-03, grad_scale: 8.0 +2023-03-21 02:31:33,209 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62334.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:31:33,550 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 02:31:38,047 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 02:31:44,018 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 02:31:55,523 INFO [train.py:901] (1/2) Epoch 23, batch 250, loss[loss=0.1349, simple_loss=0.217, pruned_loss=0.02634, over 7359.00 frames. ], tot_loss[loss=0.1468, simple_loss=0.2255, pruned_loss=0.03401, over 1032425.23 frames. ], batch size: 73, lr: 7.25e-03, grad_scale: 8.0 +2023-03-21 02:31:57,035 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 02:31:58,648 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3270, 2.3549, 2.1783, 3.4279, 1.5647, 3.4469, 1.3577, 2.8436], + device='cuda:1'), covar=tensor([0.0103, 0.1094, 0.1596, 0.0106, 0.3890, 0.0184, 0.1088, 0.0360], + device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0260, 0.0275, 0.0185, 0.0265, 0.0194, 0.0250, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:32:00,101 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62388.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:32:01,677 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62391.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:32:05,677 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62399.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:32:18,564 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 02:32:20,560 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.342e+02 1.900e+02 2.217e+02 2.652e+02 4.046e+02, threshold=4.434e+02, percent-clipped=0.0 +2023-03-21 02:32:22,053 INFO [train.py:901] (1/2) Epoch 23, batch 300, loss[loss=0.1274, simple_loss=0.2085, pruned_loss=0.02313, over 7265.00 frames. ], tot_loss[loss=0.1464, simple_loss=0.2251, pruned_loss=0.03384, over 1124367.11 frames. ], batch size: 47, lr: 7.25e-03, grad_scale: 8.0 +2023-03-21 02:32:25,534 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=62436.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:32:27,033 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 02:32:31,073 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=62447.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:32:33,695 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62452.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:32:42,415 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 +2023-03-21 02:32:46,605 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62478.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:32:47,039 INFO [train.py:901] (1/2) Epoch 23, batch 350, loss[loss=0.1201, simple_loss=0.2063, pruned_loss=0.01702, over 7349.00 frames. ], tot_loss[loss=0.146, simple_loss=0.2249, pruned_loss=0.03352, over 1196289.10 frames. ], batch size: 44, lr: 7.24e-03, grad_scale: 8.0 +2023-03-21 02:33:00,219 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 02:33:11,590 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.318e+02 1.830e+02 2.167e+02 2.718e+02 5.157e+02, threshold=4.334e+02, percent-clipped=1.0 +2023-03-21 02:33:13,145 INFO [train.py:901] (1/2) Epoch 23, batch 400, loss[loss=0.1541, simple_loss=0.2323, pruned_loss=0.03793, over 7367.00 frames. ], tot_loss[loss=0.1456, simple_loss=0.2245, pruned_loss=0.03334, over 1253014.82 frames. ], batch size: 73, lr: 7.24e-03, grad_scale: 8.0 +2023-03-21 02:33:19,214 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8018, 3.9983, 3.8476, 4.0499, 3.5965, 4.0205, 4.3149, 4.3671], + device='cuda:1'), covar=tensor([0.0230, 0.0162, 0.0202, 0.0160, 0.0291, 0.0373, 0.0218, 0.0167], + device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0114, 0.0105, 0.0110, 0.0105, 0.0094, 0.0092, 0.0088], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:33:21,322 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62545.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:33:23,868 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-03-21 02:33:29,925 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 +2023-03-21 02:33:33,719 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62570.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:33:38,736 INFO [train.py:901] (1/2) Epoch 23, batch 450, loss[loss=0.1472, simple_loss=0.2301, pruned_loss=0.03219, over 7351.00 frames. ], tot_loss[loss=0.146, simple_loss=0.2249, pruned_loss=0.03355, over 1294584.42 frames. ], batch size: 73, lr: 7.24e-03, grad_scale: 8.0 +2023-03-21 02:33:41,205 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 02:33:41,725 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 02:33:41,814 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62585.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:33:41,871 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3215, 2.9682, 3.1519, 3.0865, 2.7676, 2.5849, 3.2120, 2.5824], + device='cuda:1'), covar=tensor([0.0409, 0.0396, 0.0425, 0.0452, 0.0516, 0.0650, 0.0388, 0.1320], + device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0331, 0.0264, 0.0356, 0.0305, 0.0299, 0.0342, 0.0285], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:33:53,394 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62606.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:33:55,116 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.60 vs. limit=5.0 +2023-03-21 02:33:59,500 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=62618.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:34:01,596 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5163, 1.3139, 1.8828, 2.1469, 1.7292, 2.0977, 1.8571, 1.7800], + device='cuda:1'), covar=tensor([0.2610, 0.5137, 0.1245, 0.1262, 0.2452, 0.1387, 0.2400, 0.4559], + device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0063, 0.0050, 0.0045, 0.0046, 0.0049, 0.0073, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:34:03,454 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 1.923e+02 2.343e+02 2.764e+02 5.290e+02, threshold=4.685e+02, percent-clipped=5.0 +2023-03-21 02:34:04,974 INFO [train.py:901] (1/2) Epoch 23, batch 500, loss[loss=0.1437, simple_loss=0.2291, pruned_loss=0.02918, over 7282.00 frames. ], tot_loss[loss=0.1457, simple_loss=0.2244, pruned_loss=0.0335, over 1325834.40 frames. ], batch size: 66, lr: 7.24e-03, grad_scale: 8.0 +2023-03-21 02:34:05,061 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62629.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:34:07,158 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5119, 2.4662, 2.3280, 3.7612, 1.6998, 3.7148, 1.4266, 2.3679], + device='cuda:1'), covar=tensor([0.0158, 0.1089, 0.1688, 0.0116, 0.3717, 0.0168, 0.1187, 0.0363], + device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0260, 0.0280, 0.0187, 0.0266, 0.0194, 0.0254, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:34:13,878 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62646.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:34:16,273 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 02:34:17,784 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 02:34:18,328 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 02:34:20,324 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 02:34:24,514 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62666.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:34:24,864 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 02:34:31,467 INFO [train.py:901] (1/2) Epoch 23, batch 550, loss[loss=0.1408, simple_loss=0.2151, pruned_loss=0.03325, over 7226.00 frames. ], tot_loss[loss=0.1459, simple_loss=0.2248, pruned_loss=0.03355, over 1353112.22 frames. ], batch size: 45, lr: 7.23e-03, grad_scale: 8.0 +2023-03-21 02:34:37,023 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 02:34:44,277 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1848, 3.5780, 4.0563, 4.1920, 4.0917, 4.1616, 4.1004, 4.0320], + device='cuda:1'), covar=tensor([0.0023, 0.0091, 0.0033, 0.0022, 0.0031, 0.0028, 0.0030, 0.0039], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0057, 0.0051, 0.0047, 0.0047, 0.0051, 0.0044, 0.0061], + device='cuda:1'), out_proj_covar=tensor([7.7284e-05, 1.3115e-04, 1.0928e-04, 9.4622e-05, 9.2829e-05, 1.0227e-04, + 9.7343e-05, 1.2591e-04], device='cuda:1') +2023-03-21 02:34:45,177 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 02:34:49,162 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 02:34:55,041 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.352e+02 2.048e+02 2.314e+02 2.629e+02 3.954e+02, threshold=4.629e+02, percent-clipped=0.0 +2023-03-21 02:34:55,765 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62727.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:34:56,129 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 02:34:56,648 INFO [train.py:901] (1/2) Epoch 23, batch 600, loss[loss=0.1615, simple_loss=0.2383, pruned_loss=0.04239, over 7317.00 frames. ], tot_loss[loss=0.146, simple_loss=0.2246, pruned_loss=0.03366, over 1373957.16 frames. ], batch size: 83, lr: 7.23e-03, grad_scale: 8.0 +2023-03-21 02:35:05,751 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62747.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:35:11,490 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8281, 2.3672, 2.9738, 2.8790, 2.8143, 2.6049, 2.4844, 2.7156], + device='cuda:1'), covar=tensor([0.1871, 0.1231, 0.1237, 0.1370, 0.1042, 0.1168, 0.2090, 0.2452], + device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0057, 0.0043, 0.0044, 0.0043, 0.0040, 0.0059, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:35:13,823 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 02:35:19,970 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62773.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:35:22,436 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62778.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:35:22,851 INFO [train.py:901] (1/2) Epoch 23, batch 650, loss[loss=0.1263, simple_loss=0.2112, pruned_loss=0.02072, over 7351.00 frames. ], tot_loss[loss=0.1455, simple_loss=0.224, pruned_loss=0.03351, over 1388852.85 frames. ], batch size: 44, lr: 7.23e-03, grad_scale: 8.0 +2023-03-21 02:35:23,910 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 02:35:27,998 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6916, 2.8077, 3.5594, 3.6579, 3.6455, 3.7242, 3.5319, 3.4674], + device='cuda:1'), covar=tensor([0.0030, 0.0141, 0.0047, 0.0040, 0.0044, 0.0040, 0.0058, 0.0064], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0057, 0.0051, 0.0048, 0.0047, 0.0052, 0.0044, 0.0062], + device='cuda:1'), out_proj_covar=tensor([7.7561e-05, 1.3254e-04, 1.0996e-04, 9.5224e-05, 9.3815e-05, 1.0375e-04, + 9.7852e-05, 1.2660e-04], device='cuda:1') +2023-03-21 02:35:35,323 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9584, 2.0725, 2.0554, 2.0903, 2.3197, 2.1071, 1.7009, 1.7906], + device='cuda:1'), covar=tensor([0.0396, 0.0334, 0.0340, 0.0204, 0.0438, 0.0403, 0.0333, 0.0228], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0027, 0.0028, 0.0027, 0.0028, 0.0026, 0.0030, 0.0030], + device='cuda:1'), out_proj_covar=tensor([7.3561e-05, 7.2446e-05, 6.9694e-05, 6.7973e-05, 7.1890e-05, 6.7906e-05, + 7.4631e-05, 7.6183e-05], device='cuda:1') +2023-03-21 02:35:40,835 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 02:35:46,854 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.383e+02 1.878e+02 2.197e+02 2.481e+02 4.239e+02, threshold=4.395e+02, percent-clipped=0.0 +2023-03-21 02:35:46,939 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=62826.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:35:48,391 INFO [train.py:901] (1/2) Epoch 23, batch 700, loss[loss=0.1344, simple_loss=0.2132, pruned_loss=0.02784, over 7328.00 frames. ], tot_loss[loss=0.1458, simple_loss=0.2243, pruned_loss=0.03367, over 1398224.83 frames. ], batch size: 83, lr: 7.22e-03, grad_scale: 8.0 +2023-03-21 02:35:48,887 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 02:35:51,040 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62834.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:35:58,420 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 +2023-03-21 02:36:12,744 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 02:36:13,268 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 02:36:14,752 INFO [train.py:901] (1/2) Epoch 23, batch 750, loss[loss=0.1569, simple_loss=0.2416, pruned_loss=0.03612, over 7170.00 frames. ], tot_loss[loss=0.1455, simple_loss=0.224, pruned_loss=0.03357, over 1406169.52 frames. ], batch size: 98, lr: 7.22e-03, grad_scale: 8.0 +2023-03-21 02:36:21,885 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9853, 1.8252, 2.3289, 2.3709, 2.0807, 2.3881, 2.2224, 2.2525], + device='cuda:1'), covar=tensor([0.1788, 0.2795, 0.1312, 0.1488, 0.1995, 0.2739, 0.1637, 0.2132], + device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0064, 0.0050, 0.0045, 0.0048, 0.0049, 0.0073, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:36:25,844 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62901.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:36:26,728 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 02:36:31,720 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 02:36:37,178 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 02:36:38,151 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.391e+02 1.862e+02 2.148e+02 2.738e+02 4.952e+02, threshold=4.297e+02, percent-clipped=2.0 +2023-03-21 02:36:38,197 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 02:36:40,261 INFO [train.py:901] (1/2) Epoch 23, batch 800, loss[loss=0.1328, simple_loss=0.1985, pruned_loss=0.03353, over 7010.00 frames. ], tot_loss[loss=0.146, simple_loss=0.2244, pruned_loss=0.03378, over 1413898.85 frames. ], batch size: 35, lr: 7.22e-03, grad_scale: 8.0 +2023-03-21 02:36:40,395 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62929.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:36:46,999 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62941.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:36:49,434 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 02:36:51,526 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5729, 1.2815, 1.8739, 2.0568, 1.6980, 2.1234, 1.5735, 1.8102], + device='cuda:1'), covar=tensor([0.2044, 0.3791, 0.1513, 0.0818, 0.3281, 0.1800, 0.1834, 0.3698], + device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0064, 0.0051, 0.0045, 0.0048, 0.0049, 0.0073, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:37:04,838 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=62977.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:37:05,828 INFO [train.py:901] (1/2) Epoch 23, batch 850, loss[loss=0.1474, simple_loss=0.2258, pruned_loss=0.0345, over 7289.00 frames. ], tot_loss[loss=0.1463, simple_loss=0.2248, pruned_loss=0.03386, over 1420368.86 frames. ], batch size: 70, lr: 7.22e-03, grad_scale: 8.0 +2023-03-21 02:37:08,316 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 02:37:08,325 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 02:37:14,442 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 02:37:17,840 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 02:37:18,436 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9781, 2.0811, 1.9004, 2.0160, 2.1828, 1.9695, 1.7483, 1.6620], + device='cuda:1'), covar=tensor([0.0232, 0.0297, 0.0246, 0.0109, 0.0446, 0.0354, 0.0347, 0.0298], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0028, 0.0028, 0.0026, 0.0029, 0.0027, 0.0031, 0.0031], + device='cuda:1'), out_proj_covar=tensor([7.4864e-05, 7.4345e-05, 7.1377e-05, 6.8190e-05, 7.3249e-05, 6.9358e-05, + 7.6388e-05, 7.8539e-05], device='cuda:1') +2023-03-21 02:37:29,127 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63022.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:37:31,020 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.346e+02 2.012e+02 2.274e+02 2.694e+02 4.230e+02, threshold=4.549e+02, percent-clipped=0.0 +2023-03-21 02:37:32,558 INFO [train.py:901] (1/2) Epoch 23, batch 900, loss[loss=0.1123, simple_loss=0.1911, pruned_loss=0.01678, over 7149.00 frames. ], tot_loss[loss=0.1457, simple_loss=0.2242, pruned_loss=0.03362, over 1424675.73 frames. ], batch size: 39, lr: 7.21e-03, grad_scale: 8.0 +2023-03-21 02:37:32,937 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 +2023-03-21 02:37:38,482 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.05 vs. limit=5.0 +2023-03-21 02:37:39,960 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.03 vs. limit=2.0 +2023-03-21 02:37:41,718 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63047.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:37:49,737 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-03-21 02:37:51,376 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 +2023-03-21 02:37:55,485 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 02:37:55,651 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4285, 1.7050, 1.4291, 1.6655, 1.7257, 1.6076, 1.5559, 1.1359], + device='cuda:1'), covar=tensor([0.0123, 0.0169, 0.0223, 0.0107, 0.0067, 0.0100, 0.0185, 0.0130], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0026, 0.0027, 0.0027, 0.0025, 0.0026, 0.0028, 0.0035], + device='cuda:1'), out_proj_covar=tensor([3.2026e-05, 2.9698e-05, 3.0657e-05, 2.9774e-05, 2.8928e-05, 2.8736e-05, + 3.1947e-05, 3.9617e-05], device='cuda:1') +2023-03-21 02:37:57,524 INFO [train.py:901] (1/2) Epoch 23, batch 950, loss[loss=0.1435, simple_loss=0.226, pruned_loss=0.03047, over 7324.00 frames. ], tot_loss[loss=0.1464, simple_loss=0.2249, pruned_loss=0.03391, over 1431437.60 frames. ], batch size: 59, lr: 7.21e-03, grad_scale: 8.0 +2023-03-21 02:38:03,138 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2336, 4.1661, 3.5704, 3.7153, 3.4514, 2.2986, 1.7980, 4.2701], + device='cuda:1'), covar=tensor([0.0066, 0.0065, 0.0119, 0.0085, 0.0151, 0.0586, 0.0701, 0.0063], + device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0075, 0.0097, 0.0085, 0.0111, 0.0121, 0.0122, 0.0090], + device='cuda:1'), out_proj_covar=tensor([1.1118e-04, 9.8242e-05, 1.2087e-04, 1.0946e-04, 1.3550e-04, 1.5043e-04, + 1.5340e-04, 1.0880e-04], device='cuda:1') +2023-03-21 02:38:05,575 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=63095.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:38:19,021 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 02:38:22,015 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.310e+02 2.026e+02 2.334e+02 2.791e+02 5.736e+02, threshold=4.667e+02, percent-clipped=2.0 +2023-03-21 02:38:23,496 INFO [train.py:901] (1/2) Epoch 23, batch 1000, loss[loss=0.1374, simple_loss=0.2225, pruned_loss=0.02615, over 7317.00 frames. ], tot_loss[loss=0.1468, simple_loss=0.2254, pruned_loss=0.03408, over 1434992.72 frames. ], batch size: 80, lr: 7.21e-03, grad_scale: 8.0 +2023-03-21 02:38:23,567 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63129.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:38:25,296 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-03-21 02:38:38,029 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 02:38:43,908 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 +2023-03-21 02:38:48,452 INFO [train.py:901] (1/2) Epoch 23, batch 1050, loss[loss=0.1256, simple_loss=0.1864, pruned_loss=0.0324, over 6091.00 frames. ], tot_loss[loss=0.1459, simple_loss=0.2246, pruned_loss=0.03358, over 1435603.44 frames. ], batch size: 26, lr: 7.20e-03, grad_scale: 16.0 +2023-03-21 02:39:00,309 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 02:39:01,272 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63201.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:39:02,046 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 +2023-03-21 02:39:04,617 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 02:39:11,188 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63221.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:39:13,464 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.351e+02 1.843e+02 2.146e+02 2.547e+02 4.029e+02, threshold=4.291e+02, percent-clipped=0.0 +2023-03-21 02:39:14,993 INFO [train.py:901] (1/2) Epoch 23, batch 1100, loss[loss=0.1364, simple_loss=0.2185, pruned_loss=0.02714, over 7354.00 frames. ], tot_loss[loss=0.1457, simple_loss=0.2244, pruned_loss=0.03352, over 1436699.07 frames. ], batch size: 51, lr: 7.20e-03, grad_scale: 16.0 +2023-03-21 02:39:21,040 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63241.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:39:24,942 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=63249.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:39:32,996 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 02:39:33,505 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 02:39:40,631 INFO [train.py:901] (1/2) Epoch 23, batch 1150, loss[loss=0.1482, simple_loss=0.2308, pruned_loss=0.03283, over 7114.00 frames. ], tot_loss[loss=0.1455, simple_loss=0.2241, pruned_loss=0.03343, over 1437656.41 frames. ], batch size: 98, lr: 7.20e-03, grad_scale: 8.0 +2023-03-21 02:39:42,869 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63282.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:39:46,374 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=63289.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:39:46,858 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 02:39:47,360 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 02:40:03,157 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63322.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:40:05,566 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.308e+02 1.992e+02 2.320e+02 2.694e+02 4.886e+02, threshold=4.640e+02, percent-clipped=2.0 +2023-03-21 02:40:06,565 INFO [train.py:901] (1/2) Epoch 23, batch 1200, loss[loss=0.1563, simple_loss=0.2396, pruned_loss=0.03653, over 7112.00 frames. ], tot_loss[loss=0.1452, simple_loss=0.2237, pruned_loss=0.03332, over 1436193.34 frames. ], batch size: 98, lr: 7.20e-03, grad_scale: 8.0 +2023-03-21 02:40:14,701 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2553, 4.7283, 4.7474, 4.7007, 4.6087, 4.2974, 4.7952, 4.6040], + device='cuda:1'), covar=tensor([0.0529, 0.0429, 0.0417, 0.0457, 0.0368, 0.0398, 0.0325, 0.0482], + device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0229, 0.0179, 0.0176, 0.0140, 0.0210, 0.0181, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 02:40:16,816 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9732, 2.5243, 3.0399, 2.7406, 3.1121, 2.8956, 2.4753, 2.8441], + device='cuda:1'), covar=tensor([0.1778, 0.1049, 0.1317, 0.2689, 0.1055, 0.1295, 0.2408, 0.2465], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0059, 0.0043, 0.0044, 0.0043, 0.0042, 0.0059, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:40:18,829 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7842, 3.1758, 2.5719, 3.1639, 3.1985, 2.6564, 3.0338, 2.7833], + device='cuda:1'), covar=tensor([0.0915, 0.0898, 0.1068, 0.1138, 0.0820, 0.0910, 0.1147, 0.1472], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0048, 0.0056, 0.0049, 0.0048, 0.0049, 0.0048, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:40:19,692 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 02:40:28,434 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=63370.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:40:32,891 INFO [train.py:901] (1/2) Epoch 23, batch 1250, loss[loss=0.1359, simple_loss=0.2098, pruned_loss=0.03097, over 7147.00 frames. ], tot_loss[loss=0.1461, simple_loss=0.2244, pruned_loss=0.03388, over 1436332.49 frames. ], batch size: 39, lr: 7.19e-03, grad_scale: 8.0 +2023-03-21 02:40:35,768 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.41 vs. limit=5.0 +2023-03-21 02:40:44,309 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 02:40:48,317 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 02:40:49,752 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 02:40:53,794 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63420.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:40:54,801 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63422.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:40:56,804 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3882, 4.3384, 3.8514, 3.8163, 3.6407, 2.3998, 1.8277, 4.3862], + device='cuda:1'), covar=tensor([0.0034, 0.0031, 0.0068, 0.0051, 0.0091, 0.0432, 0.0538, 0.0040], + device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0076, 0.0097, 0.0085, 0.0112, 0.0122, 0.0121, 0.0091], + device='cuda:1'), out_proj_covar=tensor([1.1208e-04, 9.9680e-05, 1.2074e-04, 1.0938e-04, 1.3619e-04, 1.5089e-04, + 1.5307e-04, 1.1020e-04], device='cuda:1') +2023-03-21 02:40:57,169 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.231e+02 1.941e+02 2.406e+02 2.813e+02 7.014e+02, threshold=4.811e+02, percent-clipped=2.0 +2023-03-21 02:40:58,169 INFO [train.py:901] (1/2) Epoch 23, batch 1300, loss[loss=0.1567, simple_loss=0.2356, pruned_loss=0.03891, over 7323.00 frames. ], tot_loss[loss=0.1463, simple_loss=0.2245, pruned_loss=0.03403, over 1436907.96 frames. ], batch size: 61, lr: 7.19e-03, grad_scale: 8.0 +2023-03-21 02:40:58,286 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63429.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:41:02,406 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0666, 2.7832, 2.0170, 2.9577, 2.8141, 3.2074, 2.6063, 2.5625], + device='cuda:1'), covar=tensor([0.1975, 0.0698, 0.3175, 0.0600, 0.0118, 0.0173, 0.0252, 0.0234], + device='cuda:1'), in_proj_covar=tensor([0.0267, 0.0235, 0.0262, 0.0264, 0.0170, 0.0171, 0.0201, 0.0212], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:41:14,490 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 02:41:16,539 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 02:41:20,037 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 02:41:23,536 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=63477.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:41:24,441 INFO [train.py:901] (1/2) Epoch 23, batch 1350, loss[loss=0.1727, simple_loss=0.2421, pruned_loss=0.0517, over 7330.00 frames. ], tot_loss[loss=0.1461, simple_loss=0.2243, pruned_loss=0.03388, over 1436746.01 frames. ], batch size: 75, lr: 7.19e-03, grad_scale: 8.0 +2023-03-21 02:41:25,587 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63481.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:41:26,570 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63483.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:41:30,046 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 02:41:33,162 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5796, 3.7920, 3.6358, 3.7885, 3.5195, 3.7712, 4.0090, 4.0580], + device='cuda:1'), covar=tensor([0.0244, 0.0179, 0.0222, 0.0177, 0.0364, 0.0269, 0.0234, 0.0180], + device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0114, 0.0106, 0.0111, 0.0105, 0.0094, 0.0095, 0.0089], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:41:44,764 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9395, 4.0701, 3.0414, 4.1285, 3.6858, 3.9377, 2.5928, 2.6958], + device='cuda:1'), covar=tensor([0.0353, 0.0712, 0.1787, 0.0392, 0.0460, 0.0733, 0.2414, 0.1979], + device='cuda:1'), in_proj_covar=tensor([0.0265, 0.0250, 0.0285, 0.0260, 0.0265, 0.0257, 0.0249, 0.0269], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:41:48,584 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.437e+02 1.965e+02 2.363e+02 2.855e+02 5.705e+02, threshold=4.726e+02, percent-clipped=1.0 +2023-03-21 02:41:49,579 INFO [train.py:901] (1/2) Epoch 23, batch 1400, loss[loss=0.1146, simple_loss=0.1863, pruned_loss=0.02148, over 7199.00 frames. ], tot_loss[loss=0.1455, simple_loss=0.2242, pruned_loss=0.03342, over 1437609.53 frames. ], batch size: 39, lr: 7.18e-03, grad_scale: 8.0 +2023-03-21 02:42:03,243 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 02:42:08,157 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-21 02:42:10,921 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63569.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:42:14,893 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63577.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:42:15,835 INFO [train.py:901] (1/2) Epoch 23, batch 1450, loss[loss=0.1475, simple_loss=0.227, pruned_loss=0.03398, over 7329.00 frames. ], tot_loss[loss=0.1451, simple_loss=0.2238, pruned_loss=0.03326, over 1438949.75 frames. ], batch size: 54, lr: 7.18e-03, grad_scale: 8.0 +2023-03-21 02:42:19,376 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63586.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:42:25,441 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0289, 2.6744, 3.0783, 2.9477, 2.6944, 2.4053, 2.9700, 2.4051], + device='cuda:1'), covar=tensor([0.0364, 0.0481, 0.0457, 0.0425, 0.0503, 0.0737, 0.0630, 0.1524], + device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0333, 0.0266, 0.0358, 0.0308, 0.0302, 0.0343, 0.0286], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:42:25,792 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 02:42:40,736 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.192e+02 1.916e+02 2.282e+02 2.775e+02 5.456e+02, threshold=4.565e+02, percent-clipped=1.0 +2023-03-21 02:42:42,326 INFO [train.py:901] (1/2) Epoch 23, batch 1500, loss[loss=0.1574, simple_loss=0.2371, pruned_loss=0.03882, over 7137.00 frames. ], tot_loss[loss=0.1455, simple_loss=0.2241, pruned_loss=0.03349, over 1439723.04 frames. ], batch size: 98, lr: 7.18e-03, grad_scale: 8.0 +2023-03-21 02:42:42,864 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 02:42:42,995 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63630.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:42:51,547 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63647.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:43:07,532 INFO [train.py:901] (1/2) Epoch 23, batch 1550, loss[loss=0.1529, simple_loss=0.2289, pruned_loss=0.03844, over 7325.00 frames. ], tot_loss[loss=0.1446, simple_loss=0.2231, pruned_loss=0.03302, over 1439597.97 frames. ], batch size: 61, lr: 7.18e-03, grad_scale: 8.0 +2023-03-21 02:43:07,565 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 02:43:32,997 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.430e+02 1.854e+02 2.117e+02 2.642e+02 5.198e+02, threshold=4.235e+02, percent-clipped=1.0 +2023-03-21 02:43:33,933 INFO [train.py:901] (1/2) Epoch 23, batch 1600, loss[loss=0.1677, simple_loss=0.2406, pruned_loss=0.04737, over 7273.00 frames. ], tot_loss[loss=0.1449, simple_loss=0.2234, pruned_loss=0.03317, over 1440200.64 frames. ], batch size: 57, lr: 7.17e-03, grad_scale: 8.0 +2023-03-21 02:43:40,490 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 02:43:40,980 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 02:43:44,008 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 02:43:54,154 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 02:43:57,673 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 02:43:57,725 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63776.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:43:58,736 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63778.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:43:59,172 INFO [train.py:901] (1/2) Epoch 23, batch 1650, loss[loss=0.1569, simple_loss=0.2272, pruned_loss=0.04326, over 7329.00 frames. ], tot_loss[loss=0.1446, simple_loss=0.2231, pruned_loss=0.033, over 1441283.42 frames. ], batch size: 61, lr: 7.17e-03, grad_scale: 8.0 +2023-03-21 02:44:03,381 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0933, 2.2318, 2.0661, 3.5492, 1.5545, 3.1511, 1.2660, 2.8708], + device='cuda:1'), covar=tensor([0.0138, 0.1245, 0.1633, 0.0126, 0.3943, 0.0114, 0.1227, 0.0275], + device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0261, 0.0280, 0.0190, 0.0265, 0.0197, 0.0255, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:44:05,650 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 02:44:14,892 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7133, 2.1510, 1.9313, 2.0271, 2.0905, 1.8598, 1.6541, 1.7148], + device='cuda:1'), covar=tensor([0.0211, 0.0106, 0.0147, 0.0152, 0.0085, 0.0094, 0.0126, 0.0111], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0027, 0.0027, 0.0027, 0.0026, 0.0026, 0.0028, 0.0035], + device='cuda:1'), out_proj_covar=tensor([3.2649e-05, 3.0182e-05, 3.0707e-05, 3.0442e-05, 2.9300e-05, 2.9195e-05, + 3.1739e-05, 3.9771e-05], device='cuda:1') +2023-03-21 02:44:16,388 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63810.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:44:24,297 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 02:44:24,706 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.358e+02 1.919e+02 2.326e+02 2.793e+02 7.280e+02, threshold=4.653e+02, percent-clipped=3.0 +2023-03-21 02:44:25,739 INFO [train.py:901] (1/2) Epoch 23, batch 1700, loss[loss=0.1525, simple_loss=0.2347, pruned_loss=0.03513, over 7322.00 frames. ], tot_loss[loss=0.1444, simple_loss=0.2229, pruned_loss=0.0329, over 1438958.74 frames. ], batch size: 59, lr: 7.17e-03, grad_scale: 8.0 +2023-03-21 02:44:28,346 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 02:44:33,943 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63845.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:44:39,474 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 02:44:47,223 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63871.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:44:50,232 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63877.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:44:51,128 INFO [train.py:901] (1/2) Epoch 23, batch 1750, loss[loss=0.1461, simple_loss=0.2241, pruned_loss=0.03405, over 7306.00 frames. ], tot_loss[loss=0.144, simple_loss=0.2226, pruned_loss=0.0327, over 1440394.22 frames. ], batch size: 86, lr: 7.16e-03, grad_scale: 8.0 +2023-03-21 02:44:55,303 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9114, 2.6550, 2.9474, 2.9778, 2.9609, 2.8522, 2.5699, 3.1321], + device='cuda:1'), covar=tensor([0.1717, 0.0878, 0.1603, 0.1845, 0.1030, 0.0949, 0.2281, 0.1167], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0058, 0.0044, 0.0043, 0.0043, 0.0042, 0.0059, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:45:03,779 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 02:45:04,768 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 02:45:05,904 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63906.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:45:11,082 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63916.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:45:15,464 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:45:15,479 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:45:16,389 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 2.046e+02 2.411e+02 2.952e+02 4.779e+02, threshold=4.822e+02, percent-clipped=2.0 +2023-03-21 02:45:17,410 INFO [train.py:901] (1/2) Epoch 23, batch 1800, loss[loss=0.1362, simple_loss=0.22, pruned_loss=0.02623, over 7272.00 frames. ], tot_loss[loss=0.1449, simple_loss=0.2233, pruned_loss=0.03327, over 1439551.43 frames. ], batch size: 47, lr: 7.16e-03, grad_scale: 8.0 +2023-03-21 02:45:24,031 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63942.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:45:25,506 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 02:45:37,598 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 02:45:42,973 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63977.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:45:43,822 INFO [train.py:901] (1/2) Epoch 23, batch 1850, loss[loss=0.1114, simple_loss=0.1834, pruned_loss=0.01969, over 6973.00 frames. ], tot_loss[loss=0.1455, simple_loss=0.2244, pruned_loss=0.03335, over 1439383.10 frames. ], batch size: 35, lr: 7.16e-03, grad_scale: 8.0 +2023-03-21 02:45:48,591 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 02:45:49,986 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 +2023-03-21 02:46:08,854 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 02:46:09,940 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64023.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:46:11,804 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.904e+02 2.192e+02 2.514e+02 3.885e+02, threshold=4.384e+02, percent-clipped=0.0 +2023-03-21 02:46:12,806 INFO [train.py:901] (1/2) Epoch 23, batch 1900, loss[loss=0.141, simple_loss=0.223, pruned_loss=0.02947, over 7266.00 frames. ], tot_loss[loss=0.1463, simple_loss=0.2252, pruned_loss=0.03368, over 1441062.76 frames. ], batch size: 70, lr: 7.16e-03, grad_scale: 8.0 +2023-03-21 02:46:34,942 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 02:46:37,573 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64076.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:46:38,585 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64078.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:46:38,975 INFO [train.py:901] (1/2) Epoch 23, batch 1950, loss[loss=0.1438, simple_loss=0.2259, pruned_loss=0.03081, over 7269.00 frames. ], tot_loss[loss=0.1461, simple_loss=0.225, pruned_loss=0.03367, over 1442179.09 frames. ], batch size: 47, lr: 7.15e-03, grad_scale: 8.0 +2023-03-21 02:46:41,757 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64084.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:46:46,676 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 02:46:51,181 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 02:46:51,617 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 02:46:52,997 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 +2023-03-21 02:47:01,743 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=64124.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:47:02,742 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=64126.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:47:03,155 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.380e+02 1.828e+02 2.267e+02 2.905e+02 5.773e+02, threshold=4.533e+02, percent-clipped=4.0 +2023-03-21 02:47:04,191 INFO [train.py:901] (1/2) Epoch 23, batch 2000, loss[loss=0.1391, simple_loss=0.2155, pruned_loss=0.03133, over 7338.00 frames. ], tot_loss[loss=0.147, simple_loss=0.2257, pruned_loss=0.03416, over 1442979.12 frames. ], batch size: 54, lr: 7.15e-03, grad_scale: 8.0 +2023-03-21 02:47:07,776 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 02:47:08,551 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-03-21 02:47:09,940 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2596, 2.5642, 2.1566, 3.7943, 1.7970, 3.3734, 1.5595, 2.8530], + device='cuda:1'), covar=tensor([0.0133, 0.1046, 0.1827, 0.0127, 0.3828, 0.0206, 0.1181, 0.0296], + device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0261, 0.0283, 0.0191, 0.0266, 0.0201, 0.0255, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:47:20,168 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 02:47:24,289 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64166.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:47:28,071 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 02:47:30,615 INFO [train.py:901] (1/2) Epoch 23, batch 2050, loss[loss=0.1376, simple_loss=0.2194, pruned_loss=0.02787, over 7268.00 frames. ], tot_loss[loss=0.1468, simple_loss=0.2254, pruned_loss=0.03411, over 1443484.25 frames. ], batch size: 52, lr: 7.15e-03, grad_scale: 8.0 +2023-03-21 02:47:42,271 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64201.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:47:54,369 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64225.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:47:55,248 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.315e+02 1.822e+02 2.255e+02 2.710e+02 4.829e+02, threshold=4.510e+02, percent-clipped=1.0 +2023-03-21 02:47:56,249 INFO [train.py:901] (1/2) Epoch 23, batch 2100, loss[loss=0.1375, simple_loss=0.2221, pruned_loss=0.0264, over 7293.00 frames. ], tot_loss[loss=0.1462, simple_loss=0.2248, pruned_loss=0.03384, over 1443922.20 frames. ], batch size: 86, lr: 7.15e-03, grad_scale: 8.0 +2023-03-21 02:47:57,927 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64232.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:48:02,399 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 02:48:04,141 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64242.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:48:06,032 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 02:48:19,386 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64272.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:48:19,872 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=64273.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:48:22,877 INFO [train.py:901] (1/2) Epoch 23, batch 2150, loss[loss=0.1497, simple_loss=0.2232, pruned_loss=0.03813, over 7259.00 frames. ], tot_loss[loss=0.1466, simple_loss=0.2256, pruned_loss=0.0338, over 1443602.31 frames. ], batch size: 47, lr: 7.14e-03, grad_scale: 8.0 +2023-03-21 02:48:28,558 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=64290.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:48:30,129 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64293.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:48:47,382 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.192e+02 1.910e+02 2.130e+02 2.674e+02 6.186e+02, threshold=4.260e+02, percent-clipped=3.0 +2023-03-21 02:48:48,996 INFO [train.py:901] (1/2) Epoch 23, batch 2200, loss[loss=0.1664, simple_loss=0.2412, pruned_loss=0.0458, over 7227.00 frames. ], tot_loss[loss=0.1462, simple_loss=0.225, pruned_loss=0.03373, over 1441352.04 frames. ], batch size: 93, lr: 7.14e-03, grad_scale: 8.0 +2023-03-21 02:48:52,023 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 02:48:57,135 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64345.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:49:14,223 INFO [train.py:901] (1/2) Epoch 23, batch 2250, loss[loss=0.1474, simple_loss=0.2336, pruned_loss=0.0306, over 7324.00 frames. ], tot_loss[loss=0.1452, simple_loss=0.2242, pruned_loss=0.03315, over 1443560.31 frames. ], batch size: 83, lr: 7.14e-03, grad_scale: 8.0 +2023-03-21 02:49:14,299 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64379.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:49:25,263 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7045, 2.9062, 2.4501, 4.0658, 1.8684, 3.7417, 1.4876, 2.8930], + device='cuda:1'), covar=tensor([0.0096, 0.0832, 0.1621, 0.0114, 0.3658, 0.0160, 0.1196, 0.0296], + device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0256, 0.0278, 0.0189, 0.0263, 0.0200, 0.0253, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:49:26,605 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 02:49:26,616 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 02:49:28,200 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64406.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:49:39,665 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.860e+02 2.261e+02 2.711e+02 5.624e+02, threshold=4.523e+02, percent-clipped=6.0 +2023-03-21 02:49:40,203 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 02:49:40,679 INFO [train.py:901] (1/2) Epoch 23, batch 2300, loss[loss=0.1749, simple_loss=0.251, pruned_loss=0.04945, over 7130.00 frames. ], tot_loss[loss=0.1456, simple_loss=0.2245, pruned_loss=0.03334, over 1441151.08 frames. ], batch size: 98, lr: 7.13e-03, grad_scale: 8.0 +2023-03-21 02:49:58,971 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64465.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:49:59,464 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64466.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:50:05,796 INFO [train.py:901] (1/2) Epoch 23, batch 2350, loss[loss=0.1467, simple_loss=0.2285, pruned_loss=0.03246, over 7269.00 frames. ], tot_loss[loss=0.1455, simple_loss=0.2245, pruned_loss=0.03329, over 1443276.12 frames. ], batch size: 77, lr: 7.13e-03, grad_scale: 8.0 +2023-03-21 02:50:16,990 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64501.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:50:24,002 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=64514.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:50:26,504 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 02:50:30,223 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64526.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:50:30,556 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 1.893e+02 2.267e+02 2.617e+02 7.121e+02, threshold=4.533e+02, percent-clipped=2.0 +2023-03-21 02:50:31,569 INFO [train.py:901] (1/2) Epoch 23, batch 2400, loss[loss=0.1498, simple_loss=0.2277, pruned_loss=0.03597, over 7288.00 frames. ], tot_loss[loss=0.1454, simple_loss=0.2244, pruned_loss=0.03326, over 1442306.69 frames. ], batch size: 66, lr: 7.13e-03, grad_scale: 8.0 +2023-03-21 02:50:33,556 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 02:50:40,258 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8272, 3.1768, 3.9479, 3.7120, 3.9131, 3.9901, 3.9709, 3.8150], + device='cuda:1'), covar=tensor([0.0033, 0.0111, 0.0031, 0.0041, 0.0030, 0.0026, 0.0035, 0.0046], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0059, 0.0052, 0.0049, 0.0049, 0.0053, 0.0046, 0.0064], + device='cuda:1'), out_proj_covar=tensor([8.0785e-05, 1.3598e-04, 1.1183e-04, 9.7496e-05, 9.4544e-05, 1.0408e-04, + 9.9329e-05, 1.3098e-04], device='cuda:1') +2023-03-21 02:50:41,695 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=64549.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:50:43,639 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 02:50:45,708 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 02:50:53,396 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64572.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:50:56,866 INFO [train.py:901] (1/2) Epoch 23, batch 2450, loss[loss=0.1422, simple_loss=0.2173, pruned_loss=0.03352, over 7222.00 frames. ], tot_loss[loss=0.1453, simple_loss=0.2243, pruned_loss=0.0332, over 1441743.00 frames. ], batch size: 45, lr: 7.13e-03, grad_scale: 8.0 +2023-03-21 02:51:02,795 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64588.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:51:03,329 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64589.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:51:12,618 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5712, 2.9446, 2.3644, 3.8919, 1.8080, 3.6667, 1.4771, 2.8386], + device='cuda:1'), covar=tensor([0.0102, 0.0893, 0.1867, 0.0151, 0.4258, 0.0169, 0.1265, 0.0339], + device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0257, 0.0278, 0.0189, 0.0264, 0.0201, 0.0252, 0.0229], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:51:12,987 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 02:51:19,111 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=64620.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:51:22,541 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.432e+02 1.913e+02 2.327e+02 2.800e+02 7.811e+02, threshold=4.653e+02, percent-clipped=3.0 +2023-03-21 02:51:23,554 INFO [train.py:901] (1/2) Epoch 23, batch 2500, loss[loss=0.1681, simple_loss=0.2363, pruned_loss=0.04994, over 7219.00 frames. ], tot_loss[loss=0.1455, simple_loss=0.2245, pruned_loss=0.03326, over 1441742.54 frames. ], batch size: 45, lr: 7.12e-03, grad_scale: 8.0 +2023-03-21 02:51:24,721 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5364, 3.3321, 2.4748, 4.1246, 2.8105, 3.0805, 1.8575, 2.3659], + device='cuda:1'), covar=tensor([0.0342, 0.0671, 0.2112, 0.0389, 0.0332, 0.0488, 0.2884, 0.1586], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0256, 0.0293, 0.0264, 0.0272, 0.0264, 0.0255, 0.0274], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 02:51:33,230 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64648.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:51:34,321 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64650.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:51:38,856 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4219, 3.5883, 3.3547, 3.5162, 3.3419, 3.4642, 3.7644, 3.8281], + device='cuda:1'), covar=tensor([0.0258, 0.0202, 0.0300, 0.0210, 0.0348, 0.0495, 0.0260, 0.0198], + device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0116, 0.0107, 0.0111, 0.0106, 0.0094, 0.0094, 0.0090], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:51:39,252 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 02:51:49,991 INFO [train.py:901] (1/2) Epoch 23, batch 2550, loss[loss=0.1543, simple_loss=0.2398, pruned_loss=0.03439, over 7305.00 frames. ], tot_loss[loss=0.1454, simple_loss=0.2244, pruned_loss=0.03323, over 1442122.83 frames. ], batch size: 80, lr: 7.12e-03, grad_scale: 8.0 +2023-03-21 02:51:50,128 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64679.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:52:01,117 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64701.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:52:02,167 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64703.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:52:05,183 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64709.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:52:14,058 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.258e+02 1.943e+02 2.289e+02 2.838e+02 6.373e+02, threshold=4.577e+02, percent-clipped=4.0 +2023-03-21 02:52:14,132 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=64727.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:52:15,076 INFO [train.py:901] (1/2) Epoch 23, batch 2600, loss[loss=0.1668, simple_loss=0.2416, pruned_loss=0.04604, over 7342.00 frames. ], tot_loss[loss=0.1454, simple_loss=0.2244, pruned_loss=0.03318, over 1441579.84 frames. ], batch size: 54, lr: 7.12e-03, grad_scale: 8.0 +2023-03-21 02:52:20,694 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4436, 1.0226, 1.5369, 1.8308, 1.6238, 1.7740, 1.4124, 1.7296], + device='cuda:1'), covar=tensor([0.1779, 0.2821, 0.1304, 0.0870, 0.1559, 0.2612, 0.1822, 0.2140], + device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0064, 0.0051, 0.0046, 0.0048, 0.0048, 0.0076, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:52:23,154 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64745.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:52:26,611 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64752.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:52:32,518 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64764.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:52:39,675 INFO [train.py:901] (1/2) Epoch 23, batch 2650, loss[loss=0.1457, simple_loss=0.2242, pruned_loss=0.03361, over 7274.00 frames. ], tot_loss[loss=0.146, simple_loss=0.2249, pruned_loss=0.03357, over 1442018.56 frames. ], batch size: 70, lr: 7.11e-03, grad_scale: 8.0 +2023-03-21 02:52:40,990 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-03-21 02:52:53,596 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64806.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:52:54,529 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4275, 2.7361, 2.4363, 2.7443, 2.6262, 2.4355, 2.7114, 2.4284], + device='cuda:1'), covar=tensor([0.0979, 0.0691, 0.1305, 0.0809, 0.1019, 0.1173, 0.0731, 0.1844], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0050, 0.0058, 0.0051, 0.0050, 0.0051, 0.0050, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:52:56,951 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64813.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:52:57,476 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0450, 2.8890, 1.9324, 3.3420, 2.0948, 2.7755, 1.4498, 1.8287], + device='cuda:1'), covar=tensor([0.0443, 0.0811, 0.2527, 0.0785, 0.0616, 0.0532, 0.3371, 0.1933], + device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0254, 0.0291, 0.0262, 0.0271, 0.0261, 0.0253, 0.0273], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 02:53:01,342 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64821.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:53:04,338 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.370e+02 2.063e+02 2.379e+02 2.895e+02 7.875e+02, threshold=4.758e+02, percent-clipped=6.0 +2023-03-21 02:53:05,364 INFO [train.py:901] (1/2) Epoch 23, batch 2700, loss[loss=0.1574, simple_loss=0.2388, pruned_loss=0.03797, over 7331.00 frames. ], tot_loss[loss=0.1454, simple_loss=0.2245, pruned_loss=0.03311, over 1441830.70 frames. ], batch size: 61, lr: 7.11e-03, grad_scale: 8.0 +2023-03-21 02:53:09,904 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9065, 3.6736, 4.0800, 3.9577, 4.0072, 4.0955, 4.1443, 4.0071], + device='cuda:1'), covar=tensor([0.0035, 0.0086, 0.0034, 0.0046, 0.0033, 0.0029, 0.0031, 0.0055], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0058, 0.0051, 0.0048, 0.0048, 0.0052, 0.0045, 0.0063], + device='cuda:1'), out_proj_covar=tensor([7.8996e-05, 1.3382e-04, 1.0990e-04, 9.4946e-05, 9.2287e-05, 1.0211e-04, + 9.7976e-05, 1.2828e-04], device='cuda:1') +2023-03-21 02:53:15,813 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0235, 3.2631, 2.7692, 3.4106, 3.0848, 2.7650, 3.2528, 2.8855], + device='cuda:1'), covar=tensor([0.0593, 0.0782, 0.0790, 0.0520, 0.1116, 0.0946, 0.0897, 0.0962], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0049, 0.0057, 0.0050, 0.0049, 0.0051, 0.0049, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:53:26,063 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3931, 4.8992, 4.9963, 4.9208, 4.7193, 4.4768, 5.0099, 4.8218], + device='cuda:1'), covar=tensor([0.0498, 0.0391, 0.0308, 0.0437, 0.0353, 0.0360, 0.0308, 0.0407], + device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0228, 0.0175, 0.0175, 0.0140, 0.0207, 0.0178, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 02:53:30,378 INFO [train.py:901] (1/2) Epoch 23, batch 2750, loss[loss=0.1263, simple_loss=0.1974, pruned_loss=0.02758, over 7195.00 frames. ], tot_loss[loss=0.1446, simple_loss=0.2238, pruned_loss=0.03269, over 1441976.02 frames. ], batch size: 39, lr: 7.11e-03, grad_scale: 8.0 +2023-03-21 02:53:34,924 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64888.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:53:36,884 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8823, 2.5565, 3.2656, 3.0579, 3.0273, 2.6968, 2.4170, 2.8925], + device='cuda:1'), covar=tensor([0.2014, 0.0909, 0.0899, 0.1024, 0.1159, 0.1593, 0.2232, 0.1433], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0058, 0.0043, 0.0044, 0.0044, 0.0042, 0.0060, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:53:38,382 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7195, 3.9474, 3.7490, 3.9096, 3.5723, 3.8299, 4.1118, 4.1771], + device='cuda:1'), covar=tensor([0.0227, 0.0148, 0.0203, 0.0166, 0.0309, 0.0400, 0.0223, 0.0168], + device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0115, 0.0106, 0.0110, 0.0105, 0.0095, 0.0093, 0.0089], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:53:54,087 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.330e+02 1.905e+02 2.247e+02 2.674e+02 6.835e+02, threshold=4.495e+02, percent-clipped=3.0 +2023-03-21 02:53:54,960 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.74 vs. limit=5.0 +2023-03-21 02:53:55,113 INFO [train.py:901] (1/2) Epoch 23, batch 2800, loss[loss=0.1347, simple_loss=0.2102, pruned_loss=0.02961, over 7298.00 frames. ], tot_loss[loss=0.1451, simple_loss=0.2242, pruned_loss=0.03297, over 1442145.54 frames. ], batch size: 86, lr: 7.11e-03, grad_scale: 8.0 +2023-03-21 02:53:58,579 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=64936.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:54:02,980 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64945.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:54:20,167 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 02:54:20,220 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 02:54:20,245 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 02:54:20,262 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 02:54:20,432 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 02:54:20,448 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 02:54:20,459 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 02:54:20,556 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 02:54:20,630 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 02:54:20,645 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 02:54:20,989 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 02:54:21,149 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 02:54:21,154 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 02:54:21,161 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 02:54:21,301 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 02:54:21,304 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 02:54:21,361 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 02:54:29,257 INFO [train.py:901] (1/2) Epoch 24, batch 0, loss[loss=0.1609, simple_loss=0.2266, pruned_loss=0.04759, over 7242.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2266, pruned_loss=0.04759, over 7242.00 frames. ], batch size: 45, lr: 6.96e-03, grad_scale: 8.0 +2023-03-21 02:54:29,257 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 02:54:43,973 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2106, 4.1872, 3.5948, 3.6440, 3.6851, 2.4912, 2.0798, 4.2305], + device='cuda:1'), covar=tensor([0.0040, 0.0050, 0.0066, 0.0057, 0.0064, 0.0508, 0.0594, 0.0043], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0078, 0.0098, 0.0085, 0.0113, 0.0125, 0.0122, 0.0091], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 02:54:53,088 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5197, 1.1981, 1.7307, 1.9412, 1.6583, 1.8819, 1.7546, 1.7979], + device='cuda:1'), covar=tensor([0.1505, 0.2664, 0.1509, 0.1044, 0.2765, 0.1816, 0.1901, 0.2706], + device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0063, 0.0050, 0.0045, 0.0047, 0.0048, 0.0075, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:54:53,828 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5981, 4.8653, 4.8860, 4.8569, 4.6145, 4.4897, 4.9458, 4.6469], + device='cuda:1'), covar=tensor([0.0443, 0.0321, 0.0347, 0.0460, 0.0317, 0.0298, 0.0259, 0.0429], + device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0231, 0.0177, 0.0177, 0.0140, 0.0209, 0.0178, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 02:54:55,322 INFO [train.py:935] (1/2) Epoch 24, validation: loss=0.1649, simple_loss=0.2534, pruned_loss=0.03824, over 1622729.00 frames. +2023-03-21 02:54:55,323 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 02:55:01,952 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 02:55:07,026 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64976.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:55:08,092 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2836, 3.2254, 2.0740, 3.6215, 2.4681, 3.2792, 1.6226, 1.8121], + device='cuda:1'), covar=tensor([0.0461, 0.0899, 0.2757, 0.0675, 0.0675, 0.0587, 0.3423, 0.2336], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0257, 0.0293, 0.0264, 0.0271, 0.0262, 0.0253, 0.0276], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 02:55:12,417 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 02:55:18,879 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 02:55:19,934 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65001.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:55:20,812 INFO [train.py:901] (1/2) Epoch 24, batch 50, loss[loss=0.1383, simple_loss=0.2244, pruned_loss=0.02613, over 7300.00 frames. ], tot_loss[loss=0.1473, simple_loss=0.2271, pruned_loss=0.03379, over 324813.24 frames. ], batch size: 80, lr: 6.96e-03, grad_scale: 8.0 +2023-03-21 02:55:21,396 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65004.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:55:21,849 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 02:55:24,331 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 02:55:33,338 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.354e+02 1.880e+02 2.128e+02 2.472e+02 9.567e+02, threshold=4.255e+02, percent-clipped=2.0 +2023-03-21 02:55:39,082 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65037.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 02:55:41,410 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 02:55:41,930 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 02:55:44,982 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=65049.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:55:46,890 INFO [train.py:901] (1/2) Epoch 24, batch 100, loss[loss=0.1381, simple_loss=0.211, pruned_loss=0.03263, over 7221.00 frames. ], tot_loss[loss=0.1464, simple_loss=0.2267, pruned_loss=0.0331, over 572387.15 frames. ], batch size: 45, lr: 6.95e-03, grad_scale: 8.0 +2023-03-21 02:55:48,026 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65055.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:55:49,973 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65059.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:56:11,609 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65101.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:56:12,500 INFO [train.py:901] (1/2) Epoch 24, batch 150, loss[loss=0.1321, simple_loss=0.2136, pruned_loss=0.02528, over 7219.00 frames. ], tot_loss[loss=0.1463, simple_loss=0.2259, pruned_loss=0.03334, over 765258.11 frames. ], batch size: 45, lr: 6.95e-03, grad_scale: 8.0 +2023-03-21 02:56:15,081 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65108.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:56:19,187 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65116.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:56:22,247 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65121.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:56:25,194 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.371e+02 2.098e+02 2.403e+02 2.683e+02 5.360e+02, threshold=4.806e+02, percent-clipped=1.0 +2023-03-21 02:56:38,257 INFO [train.py:901] (1/2) Epoch 24, batch 200, loss[loss=0.147, simple_loss=0.2252, pruned_loss=0.03438, over 7249.00 frames. ], tot_loss[loss=0.1458, simple_loss=0.225, pruned_loss=0.03332, over 915641.85 frames. ], batch size: 55, lr: 6.95e-03, grad_scale: 8.0 +2023-03-21 02:56:42,216 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 02:56:46,204 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 02:56:46,243 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=65169.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:56:52,502 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 02:57:00,685 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8295, 2.4656, 2.9171, 2.8036, 2.9832, 2.7480, 2.3227, 2.9526], + device='cuda:1'), covar=tensor([0.1525, 0.0976, 0.1216, 0.1487, 0.1067, 0.0998, 0.2371, 0.1303], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0059, 0.0044, 0.0045, 0.0045, 0.0043, 0.0061, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:57:03,900 INFO [train.py:901] (1/2) Epoch 24, batch 250, loss[loss=0.1404, simple_loss=0.216, pruned_loss=0.0324, over 7245.00 frames. ], tot_loss[loss=0.1463, simple_loss=0.2252, pruned_loss=0.03375, over 1034158.74 frames. ], batch size: 89, lr: 6.95e-03, grad_scale: 8.0 +2023-03-21 02:57:05,835 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 02:57:11,635 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0120, 2.4762, 1.8403, 2.8524, 2.8136, 3.0832, 2.6245, 2.3493], + device='cuda:1'), covar=tensor([0.1747, 0.0815, 0.3390, 0.0515, 0.0156, 0.0168, 0.0249, 0.0295], + device='cuda:1'), in_proj_covar=tensor([0.0264, 0.0230, 0.0261, 0.0266, 0.0174, 0.0173, 0.0200, 0.0217], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 02:57:11,860 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 +2023-03-21 02:57:16,469 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 1.997e+02 2.348e+02 3.029e+02 5.559e+02, threshold=4.697e+02, percent-clipped=4.0 +2023-03-21 02:57:18,326 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-03-21 02:57:25,751 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65245.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:57:27,125 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 02:57:29,608 INFO [train.py:901] (1/2) Epoch 24, batch 300, loss[loss=0.1246, simple_loss=0.1822, pruned_loss=0.03348, over 6069.00 frames. ], tot_loss[loss=0.1461, simple_loss=0.2248, pruned_loss=0.03372, over 1124657.91 frames. ], batch size: 26, lr: 6.94e-03, grad_scale: 8.0 +2023-03-21 02:57:35,558 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 02:57:46,250 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65285.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:57:50,774 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=65293.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:57:55,760 INFO [train.py:901] (1/2) Epoch 24, batch 350, loss[loss=0.1441, simple_loss=0.2189, pruned_loss=0.03461, over 7309.00 frames. ], tot_loss[loss=0.1444, simple_loss=0.2231, pruned_loss=0.03284, over 1194470.67 frames. ], batch size: 49, lr: 6.94e-03, grad_scale: 16.0 +2023-03-21 02:57:56,399 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65304.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:58:07,851 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.327e+02 1.918e+02 2.301e+02 2.604e+02 6.154e+02, threshold=4.603e+02, percent-clipped=3.0 +2023-03-21 02:58:09,783 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-21 02:58:10,506 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65332.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 02:58:11,894 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 02:58:17,551 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65346.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:58:20,450 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=65352.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:58:20,690 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-21 02:58:20,873 INFO [train.py:901] (1/2) Epoch 24, batch 400, loss[loss=0.1612, simple_loss=0.2355, pruned_loss=0.04344, over 7294.00 frames. ], tot_loss[loss=0.1448, simple_loss=0.2235, pruned_loss=0.03308, over 1248609.29 frames. ], batch size: 68, lr: 6.94e-03, grad_scale: 16.0 +2023-03-21 02:58:24,000 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65359.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:58:40,628 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2557, 4.2391, 3.7806, 3.6904, 3.4242, 2.3927, 2.0298, 4.3176], + device='cuda:1'), covar=tensor([0.0048, 0.0059, 0.0076, 0.0065, 0.0112, 0.0476, 0.0548, 0.0049], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0077, 0.0097, 0.0084, 0.0112, 0.0123, 0.0120, 0.0090], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 02:58:46,588 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65401.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:58:47,497 INFO [train.py:901] (1/2) Epoch 24, batch 450, loss[loss=0.1397, simple_loss=0.2202, pruned_loss=0.02965, over 7318.00 frames. ], tot_loss[loss=0.1446, simple_loss=0.2235, pruned_loss=0.03286, over 1294087.46 frames. ], batch size: 75, lr: 6.94e-03, grad_scale: 16.0 +2023-03-21 02:58:49,335 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-21 02:58:49,536 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=65407.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:58:50,067 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65408.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:58:51,566 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65411.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:58:54,025 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 02:58:54,495 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 02:58:59,519 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.265e+02 1.901e+02 2.269e+02 2.647e+02 4.135e+02, threshold=4.538e+02, percent-clipped=0.0 +2023-03-21 02:59:11,152 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=65449.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:59:13,092 INFO [train.py:901] (1/2) Epoch 24, batch 500, loss[loss=0.159, simple_loss=0.2358, pruned_loss=0.0411, over 7331.00 frames. ], tot_loss[loss=0.1454, simple_loss=0.2242, pruned_loss=0.03325, over 1326427.17 frames. ], batch size: 75, lr: 6.93e-03, grad_scale: 16.0 +2023-03-21 02:59:13,236 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9489, 2.5945, 3.1682, 2.9869, 3.1388, 2.7812, 2.5249, 2.9843], + device='cuda:1'), covar=tensor([0.1646, 0.0789, 0.0813, 0.1496, 0.0971, 0.0974, 0.2204, 0.1688], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0059, 0.0044, 0.0045, 0.0045, 0.0043, 0.0060, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 02:59:14,665 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=65456.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 02:59:26,254 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 02:59:26,592 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-21 02:59:27,747 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 02:59:28,299 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 02:59:30,242 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 02:59:34,615 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 02:59:38,603 INFO [train.py:901] (1/2) Epoch 24, batch 550, loss[loss=0.1529, simple_loss=0.2378, pruned_loss=0.03401, over 7324.00 frames. ], tot_loss[loss=0.1453, simple_loss=0.2242, pruned_loss=0.03325, over 1353065.61 frames. ], batch size: 59, lr: 6.93e-03, grad_scale: 16.0 +2023-03-21 02:59:42,191 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2237, 2.2198, 2.7058, 2.2651, 2.8459, 2.4248, 2.3937, 2.0697], + device='cuda:1'), covar=tensor([0.0258, 0.0340, 0.0302, 0.0449, 0.0322, 0.0748, 0.0202, 0.0279], + device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0030, 0.0029, 0.0028, 0.0030, 0.0029, 0.0032, 0.0032], + device='cuda:1'), out_proj_covar=tensor([7.7971e-05, 7.9151e-05, 7.3703e-05, 7.1920e-05, 7.7123e-05, 7.3747e-05, + 7.9668e-05, 8.1800e-05], device='cuda:1') +2023-03-21 02:59:46,578 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 02:59:50,481 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 1.926e+02 2.261e+02 2.748e+02 6.198e+02, threshold=4.522e+02, percent-clipped=2.0 +2023-03-21 02:59:53,175 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65532.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 02:59:54,599 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 02:59:57,596 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 03:00:04,068 INFO [train.py:901] (1/2) Epoch 24, batch 600, loss[loss=0.1392, simple_loss=0.2174, pruned_loss=0.03047, over 7268.00 frames. ], tot_loss[loss=0.1452, simple_loss=0.224, pruned_loss=0.03324, over 1371653.25 frames. ], batch size: 57, lr: 6.93e-03, grad_scale: 16.0 +2023-03-21 03:00:05,568 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 03:00:22,179 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 03:00:24,830 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65593.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 03:00:30,017 INFO [train.py:901] (1/2) Epoch 24, batch 650, loss[loss=0.1349, simple_loss=0.222, pruned_loss=0.02391, over 7281.00 frames. ], tot_loss[loss=0.1445, simple_loss=0.2237, pruned_loss=0.03269, over 1387256.39 frames. ], batch size: 57, lr: 6.93e-03, grad_scale: 16.0 +2023-03-21 03:00:30,560 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 03:00:42,616 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.867e+02 2.152e+02 2.439e+02 4.019e+02, threshold=4.303e+02, percent-clipped=0.0 +2023-03-21 03:00:45,302 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65632.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:00:47,795 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5610, 1.1569, 1.8582, 2.1447, 1.5580, 2.1465, 1.7669, 1.9089], + device='cuda:1'), covar=tensor([0.2677, 0.4606, 0.1472, 0.1348, 0.4060, 0.2375, 0.1985, 0.2189], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0062, 0.0048, 0.0044, 0.0048, 0.0046, 0.0073, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:00:48,643 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 03:00:49,733 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65641.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:00:56,267 INFO [train.py:901] (1/2) Epoch 24, batch 700, loss[loss=0.1406, simple_loss=0.2268, pruned_loss=0.02716, over 7267.00 frames. ], tot_loss[loss=0.145, simple_loss=0.2243, pruned_loss=0.03285, over 1400432.45 frames. ], batch size: 52, lr: 6.92e-03, grad_scale: 16.0 +2023-03-21 03:00:56,782 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 03:01:04,511 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3578, 3.1064, 3.1570, 3.2265, 2.7951, 2.5707, 3.4306, 2.5117], + device='cuda:1'), covar=tensor([0.0383, 0.0422, 0.0456, 0.0474, 0.0529, 0.0770, 0.0498, 0.1342], + device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0338, 0.0269, 0.0362, 0.0307, 0.0305, 0.0343, 0.0287], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:01:09,881 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=65680.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:01:15,911 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0497, 3.8249, 3.8130, 3.7604, 3.1494, 3.7091, 3.8739, 3.6467], + device='cuda:1'), covar=tensor([0.0211, 0.0195, 0.0182, 0.0201, 0.0704, 0.0158, 0.0263, 0.0220], + device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0089, 0.0089, 0.0077, 0.0156, 0.0097, 0.0092, 0.0096], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:01:19,711 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 03:01:19,727 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 03:01:21,789 INFO [train.py:901] (1/2) Epoch 24, batch 750, loss[loss=0.1445, simple_loss=0.2288, pruned_loss=0.03013, over 7257.00 frames. ], tot_loss[loss=0.1455, simple_loss=0.2248, pruned_loss=0.03309, over 1410173.47 frames. ], batch size: 64, lr: 6.92e-03, grad_scale: 16.0 +2023-03-21 03:01:25,916 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65711.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:01:33,716 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 03:01:34,193 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.194e+02 1.950e+02 2.225e+02 2.735e+02 4.700e+02, threshold=4.450e+02, percent-clipped=3.0 +2023-03-21 03:01:38,523 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3056, 2.8965, 3.2696, 3.1333, 2.8230, 2.6221, 3.3452, 2.5294], + device='cuda:1'), covar=tensor([0.0371, 0.0410, 0.0455, 0.0471, 0.0525, 0.0754, 0.0557, 0.1306], + device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0341, 0.0271, 0.0364, 0.0308, 0.0306, 0.0345, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:01:39,866 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 03:01:45,338 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 03:01:46,809 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 03:01:47,311 INFO [train.py:901] (1/2) Epoch 24, batch 800, loss[loss=0.1242, simple_loss=0.1985, pruned_loss=0.0249, over 7158.00 frames. ], tot_loss[loss=0.1451, simple_loss=0.2245, pruned_loss=0.03282, over 1418487.38 frames. ], batch size: 41, lr: 6.92e-03, grad_scale: 8.0 +2023-03-21 03:01:49,486 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5625, 3.1519, 3.4989, 3.2994, 3.0542, 2.8412, 3.5121, 2.7461], + device='cuda:1'), covar=tensor([0.0390, 0.0441, 0.0456, 0.0469, 0.0537, 0.0793, 0.0415, 0.1344], + device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0343, 0.0272, 0.0365, 0.0309, 0.0307, 0.0346, 0.0289], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:01:50,389 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=65759.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:01:57,388 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 03:02:13,257 INFO [train.py:901] (1/2) Epoch 24, batch 850, loss[loss=0.1365, simple_loss=0.2216, pruned_loss=0.0257, over 7281.00 frames. ], tot_loss[loss=0.1452, simple_loss=0.2246, pruned_loss=0.03287, over 1424059.26 frames. ], batch size: 66, lr: 6.91e-03, grad_scale: 8.0 +2023-03-21 03:02:15,687 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 03:02:15,691 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 03:02:17,557 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 +2023-03-21 03:02:21,818 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 03:02:24,861 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 03:02:26,325 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.275e+02 1.907e+02 2.307e+02 2.864e+02 5.417e+02, threshold=4.614e+02, percent-clipped=5.0 +2023-03-21 03:02:28,020 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3462, 1.5672, 1.4558, 1.4087, 1.5223, 1.5522, 1.3528, 1.2036], + device='cuda:1'), covar=tensor([0.0107, 0.0082, 0.0192, 0.0115, 0.0072, 0.0089, 0.0117, 0.0120], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0028, 0.0028, 0.0030, 0.0028, 0.0027, 0.0029, 0.0037], + device='cuda:1'), out_proj_covar=tensor([3.5193e-05, 3.1547e-05, 3.2508e-05, 3.3441e-05, 3.1989e-05, 3.0402e-05, + 3.3624e-05, 4.2629e-05], device='cuda:1') +2023-03-21 03:02:31,546 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65838.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:02:34,881 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6689, 5.2525, 5.3303, 5.2289, 4.9892, 4.7487, 5.3342, 5.1059], + device='cuda:1'), covar=tensor([0.0441, 0.0314, 0.0294, 0.0393, 0.0307, 0.0288, 0.0258, 0.0412], + device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0230, 0.0176, 0.0175, 0.0138, 0.0207, 0.0179, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 03:02:35,456 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0507, 2.0179, 2.2415, 2.3209, 2.3786, 1.9222, 1.8765, 1.6458], + device='cuda:1'), covar=tensor([0.0436, 0.0465, 0.0244, 0.0167, 0.0378, 0.0730, 0.0362, 0.0406], + device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0030, 0.0029, 0.0028, 0.0030, 0.0029, 0.0032, 0.0031], + device='cuda:1'), out_proj_covar=tensor([7.7155e-05, 7.8612e-05, 7.2922e-05, 7.1651e-05, 7.6595e-05, 7.3255e-05, + 7.9526e-05, 8.0091e-05], device='cuda:1') +2023-03-21 03:02:38,829 INFO [train.py:901] (1/2) Epoch 24, batch 900, loss[loss=0.16, simple_loss=0.2418, pruned_loss=0.03915, over 7290.00 frames. ], tot_loss[loss=0.1452, simple_loss=0.2245, pruned_loss=0.03297, over 1427288.92 frames. ], batch size: 57, lr: 6.91e-03, grad_scale: 8.0 +2023-03-21 03:02:56,804 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65888.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 03:03:02,274 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 03:03:02,406 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65899.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 03:03:04,856 INFO [train.py:901] (1/2) Epoch 24, batch 950, loss[loss=0.146, simple_loss=0.226, pruned_loss=0.03296, over 7267.00 frames. ], tot_loss[loss=0.1454, simple_loss=0.2244, pruned_loss=0.03323, over 1429196.51 frames. ], batch size: 52, lr: 6.91e-03, grad_scale: 8.0 +2023-03-21 03:03:11,479 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0369, 4.5647, 4.4013, 4.9305, 4.8666, 5.0078, 4.4677, 4.6368], + device='cuda:1'), covar=tensor([0.0840, 0.2735, 0.2560, 0.1218, 0.0985, 0.1209, 0.0795, 0.0978], + device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0345, 0.0271, 0.0268, 0.0203, 0.0330, 0.0202, 0.0243], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:03:17,338 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.335e+02 1.871e+02 2.193e+02 2.615e+02 7.204e+02, threshold=4.387e+02, percent-clipped=1.0 +2023-03-21 03:03:24,034 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65941.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:03:25,979 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 03:03:29,052 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.1258, 1.3915, 1.2309, 1.2054, 1.2597, 1.2280, 1.1960, 1.0349], + device='cuda:1'), covar=tensor([0.0166, 0.0116, 0.0245, 0.0145, 0.0098, 0.0158, 0.0123, 0.0177], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0028, 0.0028, 0.0030, 0.0028, 0.0028, 0.0030, 0.0038], + device='cuda:1'), out_proj_covar=tensor([3.5106e-05, 3.2129e-05, 3.2512e-05, 3.3725e-05, 3.1872e-05, 3.0858e-05, + 3.3942e-05, 4.2970e-05], device='cuda:1') +2023-03-21 03:03:29,898 INFO [train.py:901] (1/2) Epoch 24, batch 1000, loss[loss=0.1321, simple_loss=0.2165, pruned_loss=0.02382, over 7347.00 frames. ], tot_loss[loss=0.1447, simple_loss=0.224, pruned_loss=0.03275, over 1433479.99 frames. ], batch size: 44, lr: 6.91e-03, grad_scale: 8.0 +2023-03-21 03:03:38,017 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0389, 4.5572, 4.3690, 4.9920, 4.8859, 4.9951, 4.3615, 4.6109], + device='cuda:1'), covar=tensor([0.0894, 0.2475, 0.2819, 0.1158, 0.1003, 0.1311, 0.0891, 0.1169], + device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0347, 0.0272, 0.0269, 0.0203, 0.0333, 0.0203, 0.0245], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:03:47,318 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 03:03:47,550 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-03-21 03:03:48,851 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=65989.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:03:56,625 INFO [train.py:901] (1/2) Epoch 24, batch 1050, loss[loss=0.147, simple_loss=0.2283, pruned_loss=0.03289, over 7300.00 frames. ], tot_loss[loss=0.145, simple_loss=0.2242, pruned_loss=0.03288, over 1436760.13 frames. ], batch size: 66, lr: 6.90e-03, grad_scale: 8.0 +2023-03-21 03:04:07,427 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 +2023-03-21 03:04:09,036 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.813e+02 2.192e+02 2.509e+02 4.025e+02, threshold=4.384e+02, percent-clipped=0.0 +2023-03-21 03:04:09,595 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 03:04:13,657 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 03:04:22,113 INFO [train.py:901] (1/2) Epoch 24, batch 1100, loss[loss=0.1606, simple_loss=0.2385, pruned_loss=0.04135, over 7217.00 frames. ], tot_loss[loss=0.1456, simple_loss=0.2252, pruned_loss=0.03303, over 1438535.04 frames. ], batch size: 50, lr: 6.90e-03, grad_scale: 8.0 +2023-03-21 03:04:42,714 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 03:04:43,238 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 03:04:47,666 INFO [train.py:901] (1/2) Epoch 24, batch 1150, loss[loss=0.158, simple_loss=0.2304, pruned_loss=0.04274, over 7286.00 frames. ], tot_loss[loss=0.1453, simple_loss=0.2246, pruned_loss=0.03303, over 1437578.69 frames. ], batch size: 57, lr: 6.90e-03, grad_scale: 8.0 +2023-03-21 03:04:55,194 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 03:04:56,175 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 03:05:00,165 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.297e+02 1.914e+02 2.178e+02 2.666e+02 4.759e+02, threshold=4.356e+02, percent-clipped=2.0 +2023-03-21 03:05:09,004 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8967, 1.9399, 2.1450, 1.9324, 2.1966, 1.9415, 1.8207, 1.6060], + device='cuda:1'), covar=tensor([0.0347, 0.0530, 0.0155, 0.0174, 0.0300, 0.0352, 0.0251, 0.0404], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0030, 0.0030, 0.0028, 0.0031, 0.0030, 0.0033, 0.0032], + device='cuda:1'), out_proj_covar=tensor([7.8203e-05, 7.9082e-05, 7.4421e-05, 7.3075e-05, 7.7957e-05, 7.5225e-05, + 8.0761e-05, 8.1531e-05], device='cuda:1') +2023-03-21 03:05:09,015 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66144.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:05:13,307 INFO [train.py:901] (1/2) Epoch 24, batch 1200, loss[loss=0.14, simple_loss=0.2232, pruned_loss=0.02837, over 7283.00 frames. ], tot_loss[loss=0.1447, simple_loss=0.2238, pruned_loss=0.03278, over 1437487.65 frames. ], batch size: 70, lr: 6.90e-03, grad_scale: 8.0 +2023-03-21 03:05:16,933 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66160.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:05:29,119 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 03:05:31,698 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66188.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 03:05:34,620 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66194.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:05:38,962 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3324, 4.3807, 3.8722, 3.8634, 3.5329, 2.3343, 1.9540, 4.4121], + device='cuda:1'), covar=tensor([0.0064, 0.0047, 0.0116, 0.0063, 0.0164, 0.0617, 0.0601, 0.0063], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0077, 0.0098, 0.0082, 0.0113, 0.0120, 0.0120, 0.0090], + device='cuda:1'), out_proj_covar=tensor([1.1172e-04, 9.9913e-05, 1.2060e-04, 1.0539e-04, 1.3718e-04, 1.4858e-04, + 1.5034e-04, 1.0792e-04], device='cuda:1') +2023-03-21 03:05:39,346 INFO [train.py:901] (1/2) Epoch 24, batch 1250, loss[loss=0.1201, simple_loss=0.1931, pruned_loss=0.02359, over 7206.00 frames. ], tot_loss[loss=0.1449, simple_loss=0.2238, pruned_loss=0.03301, over 1435927.66 frames. ], batch size: 39, lr: 6.89e-03, grad_scale: 8.0 +2023-03-21 03:05:40,513 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66205.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 03:05:49,139 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66221.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 03:05:51,940 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 03:05:52,403 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.872e+02 2.228e+02 2.554e+02 4.921e+02, threshold=4.456e+02, percent-clipped=2.0 +2023-03-21 03:05:55,953 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 03:05:56,497 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=66236.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:05:56,945 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 03:06:05,559 INFO [train.py:901] (1/2) Epoch 24, batch 1300, loss[loss=0.1211, simple_loss=0.2014, pruned_loss=0.02037, over 7339.00 frames. ], tot_loss[loss=0.1453, simple_loss=0.2242, pruned_loss=0.03322, over 1439434.84 frames. ], batch size: 44, lr: 6.89e-03, grad_scale: 8.0 +2023-03-21 03:06:08,141 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66258.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:06:19,642 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 03:06:21,928 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 +2023-03-21 03:06:22,076 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 03:06:25,563 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 03:06:31,022 INFO [train.py:901] (1/2) Epoch 24, batch 1350, loss[loss=0.1496, simple_loss=0.2271, pruned_loss=0.0361, over 7289.00 frames. ], tot_loss[loss=0.1455, simple_loss=0.2248, pruned_loss=0.03309, over 1442153.67 frames. ], batch size: 70, lr: 6.89e-03, grad_scale: 8.0 +2023-03-21 03:06:36,012 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 03:06:39,152 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66319.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:06:43,444 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.222e+02 1.819e+02 2.167e+02 2.669e+02 5.155e+02, threshold=4.335e+02, percent-clipped=1.0 +2023-03-21 03:06:56,408 INFO [train.py:901] (1/2) Epoch 24, batch 1400, loss[loss=0.1583, simple_loss=0.2316, pruned_loss=0.04255, over 7333.00 frames. ], tot_loss[loss=0.1452, simple_loss=0.2246, pruned_loss=0.03288, over 1443947.49 frames. ], batch size: 61, lr: 6.89e-03, grad_scale: 8.0 +2023-03-21 03:07:08,480 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 03:07:22,312 INFO [train.py:901] (1/2) Epoch 24, batch 1450, loss[loss=0.1567, simple_loss=0.2334, pruned_loss=0.04003, over 7238.00 frames. ], tot_loss[loss=0.1446, simple_loss=0.2236, pruned_loss=0.03282, over 1442367.32 frames. ], batch size: 55, lr: 6.88e-03, grad_scale: 8.0 +2023-03-21 03:07:32,897 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 03:07:33,924 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2495, 3.6095, 4.2718, 4.1247, 4.2657, 4.3119, 4.2791, 4.0932], + device='cuda:1'), covar=tensor([0.0025, 0.0090, 0.0024, 0.0030, 0.0027, 0.0021, 0.0021, 0.0043], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0058, 0.0052, 0.0049, 0.0048, 0.0052, 0.0045, 0.0065], + device='cuda:1'), out_proj_covar=tensor([7.8628e-05, 1.3249e-04, 1.0877e-04, 9.6947e-05, 9.1979e-05, 1.0088e-04, + 9.6841e-05, 1.3103e-04], device='cuda:1') +2023-03-21 03:07:34,803 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.376e+02 1.845e+02 2.245e+02 2.730e+02 1.405e+03, threshold=4.491e+02, percent-clipped=2.0 +2023-03-21 03:07:38,101 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7563, 1.4611, 1.8165, 2.2166, 1.9304, 2.2808, 2.0586, 2.0331], + device='cuda:1'), covar=tensor([0.1500, 0.3768, 0.0977, 0.1528, 0.1772, 0.3307, 0.1947, 0.3064], + device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0062, 0.0050, 0.0046, 0.0048, 0.0049, 0.0075, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:07:38,720 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-03-21 03:07:42,125 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7754, 1.5546, 1.8528, 2.2284, 1.9848, 2.3396, 2.0632, 2.0699], + device='cuda:1'), covar=tensor([0.2301, 0.3969, 0.1495, 0.1561, 0.3718, 0.2652, 0.2307, 0.3514], + device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0062, 0.0050, 0.0046, 0.0048, 0.0049, 0.0075, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:07:48,005 INFO [train.py:901] (1/2) Epoch 24, batch 1500, loss[loss=0.1391, simple_loss=0.2172, pruned_loss=0.03056, over 7283.00 frames. ], tot_loss[loss=0.1446, simple_loss=0.2237, pruned_loss=0.03273, over 1442153.99 frames. ], batch size: 42, lr: 6.88e-03, grad_scale: 8.0 +2023-03-21 03:07:48,968 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 03:07:49,032 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2200, 4.2475, 4.0907, 4.3102, 3.9346, 4.0594, 4.4457, 4.5636], + device='cuda:1'), covar=tensor([0.0302, 0.0234, 0.0263, 0.0227, 0.0369, 0.0360, 0.0306, 0.0230], + device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0118, 0.0107, 0.0111, 0.0105, 0.0096, 0.0094, 0.0090], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:07:51,833 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 +2023-03-21 03:07:52,262 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 +2023-03-21 03:08:01,232 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7611, 3.1961, 2.6849, 3.0259, 2.9768, 2.4208, 2.7370, 2.7672], + device='cuda:1'), covar=tensor([0.0552, 0.0431, 0.0772, 0.1109, 0.0417, 0.1048, 0.1206, 0.0929], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0048, 0.0057, 0.0050, 0.0048, 0.0050, 0.0050, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:08:09,116 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66494.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 03:08:12,145 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66500.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 03:08:13,089 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 03:08:13,553 INFO [train.py:901] (1/2) Epoch 24, batch 1550, loss[loss=0.1487, simple_loss=0.2346, pruned_loss=0.0314, over 7173.00 frames. ], tot_loss[loss=0.1443, simple_loss=0.2234, pruned_loss=0.03258, over 1441658.82 frames. ], batch size: 98, lr: 6.88e-03, grad_scale: 8.0 +2023-03-21 03:08:20,765 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66516.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:08:23,226 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66521.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 03:08:26,581 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.397e+02 1.920e+02 2.387e+02 2.777e+02 7.099e+02, threshold=4.774e+02, percent-clipped=1.0 +2023-03-21 03:08:33,702 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=66542.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:08:39,089 INFO [train.py:901] (1/2) Epoch 24, batch 1600, loss[loss=0.1441, simple_loss=0.2315, pruned_loss=0.02837, over 7217.00 frames. ], tot_loss[loss=0.1447, simple_loss=0.2237, pruned_loss=0.03279, over 1442769.77 frames. ], batch size: 93, lr: 6.88e-03, grad_scale: 8.0 +2023-03-21 03:08:44,639 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 03:08:45,219 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 03:08:47,695 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 03:08:54,331 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66582.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 03:08:57,651 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 03:09:01,237 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 03:09:05,745 INFO [train.py:901] (1/2) Epoch 24, batch 1650, loss[loss=0.1467, simple_loss=0.2266, pruned_loss=0.03341, over 7265.00 frames. ], tot_loss[loss=0.1441, simple_loss=0.2233, pruned_loss=0.0324, over 1440579.66 frames. ], batch size: 47, lr: 6.87e-03, grad_scale: 8.0 +2023-03-21 03:09:10,921 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 03:09:11,500 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66614.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:09:18,401 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.348e+02 1.874e+02 2.286e+02 2.733e+02 3.925e+02, threshold=4.573e+02, percent-clipped=0.0 +2023-03-21 03:09:19,100 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0206, 2.8671, 1.9599, 3.3068, 2.0330, 2.8331, 1.4083, 1.9039], + device='cuda:1'), covar=tensor([0.0466, 0.0714, 0.2674, 0.0667, 0.0477, 0.0554, 0.3329, 0.1793], + device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0253, 0.0291, 0.0265, 0.0272, 0.0261, 0.0255, 0.0273], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 03:09:27,987 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 03:09:31,669 INFO [train.py:901] (1/2) Epoch 24, batch 1700, loss[loss=0.1498, simple_loss=0.2295, pruned_loss=0.03501, over 7285.00 frames. ], tot_loss[loss=0.144, simple_loss=0.2232, pruned_loss=0.0324, over 1440163.95 frames. ], batch size: 70, lr: 6.87e-03, grad_scale: 8.0 +2023-03-21 03:09:33,229 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 03:09:43,371 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 03:09:57,424 INFO [train.py:901] (1/2) Epoch 24, batch 1750, loss[loss=0.1638, simple_loss=0.2397, pruned_loss=0.04394, over 7143.00 frames. ], tot_loss[loss=0.1436, simple_loss=0.2226, pruned_loss=0.03228, over 1440675.24 frames. ], batch size: 98, lr: 6.87e-03, grad_scale: 8.0 +2023-03-21 03:10:05,729 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 +2023-03-21 03:10:06,939 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 03:10:08,415 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 03:10:09,892 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.269e+02 1.848e+02 2.142e+02 2.647e+02 7.442e+02, threshold=4.285e+02, percent-clipped=1.0 +2023-03-21 03:10:12,223 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66731.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:10:23,250 INFO [train.py:901] (1/2) Epoch 24, batch 1800, loss[loss=0.1339, simple_loss=0.2071, pruned_loss=0.03041, over 7172.00 frames. ], tot_loss[loss=0.1436, simple_loss=0.2227, pruned_loss=0.03226, over 1441450.71 frames. ], batch size: 39, lr: 6.87e-03, grad_scale: 8.0 +2023-03-21 03:10:30,754 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 03:10:43,468 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66792.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:10:44,471 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0580, 2.0640, 2.1478, 2.2372, 2.3420, 1.9472, 1.9362, 1.7637], + device='cuda:1'), covar=tensor([0.0246, 0.0420, 0.0200, 0.0168, 0.0546, 0.0690, 0.0296, 0.0248], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0031, 0.0030, 0.0029, 0.0031, 0.0030, 0.0034, 0.0032], + device='cuda:1'), out_proj_covar=tensor([7.9485e-05, 8.0353e-05, 7.5212e-05, 7.4724e-05, 7.9238e-05, 7.5978e-05, + 8.2086e-05, 8.2577e-05], device='cuda:1') +2023-03-21 03:10:44,864 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 03:10:47,764 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66800.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 03:10:49,123 INFO [train.py:901] (1/2) Epoch 24, batch 1850, loss[loss=0.1568, simple_loss=0.2351, pruned_loss=0.03925, over 7263.00 frames. ], tot_loss[loss=0.144, simple_loss=0.223, pruned_loss=0.03252, over 1443640.72 frames. ], batch size: 52, lr: 6.86e-03, grad_scale: 8.0 +2023-03-21 03:10:55,728 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 03:10:55,804 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66816.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:11:02,369 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.480e+02 1.896e+02 2.287e+02 2.722e+02 5.764e+02, threshold=4.573e+02, percent-clipped=2.0 +2023-03-21 03:11:12,509 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=66848.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:11:13,481 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 03:11:15,003 INFO [train.py:901] (1/2) Epoch 24, batch 1900, loss[loss=0.132, simple_loss=0.2123, pruned_loss=0.02582, over 7318.00 frames. ], tot_loss[loss=0.1436, simple_loss=0.2223, pruned_loss=0.03248, over 1438010.11 frames. ], batch size: 83, lr: 6.86e-03, grad_scale: 8.0 +2023-03-21 03:11:20,532 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=66864.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 03:11:27,510 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66877.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:11:37,204 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 03:11:37,820 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 03:11:40,321 INFO [train.py:901] (1/2) Epoch 24, batch 1950, loss[loss=0.1336, simple_loss=0.2129, pruned_loss=0.02719, over 7308.00 frames. ], tot_loss[loss=0.1436, simple_loss=0.2222, pruned_loss=0.03244, over 1437490.23 frames. ], batch size: 49, lr: 6.86e-03, grad_scale: 8.0 +2023-03-21 03:11:46,453 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66914.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:11:48,809 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 03:11:53,316 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.231e+02 1.892e+02 2.217e+02 2.546e+02 5.522e+02, threshold=4.435e+02, percent-clipped=0.0 +2023-03-21 03:11:53,344 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 03:11:53,866 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 03:12:05,968 INFO [train.py:901] (1/2) Epoch 24, batch 2000, loss[loss=0.1408, simple_loss=0.2247, pruned_loss=0.02845, over 7244.00 frames. ], tot_loss[loss=0.1438, simple_loss=0.2231, pruned_loss=0.03221, over 1439194.59 frames. ], batch size: 89, lr: 6.86e-03, grad_scale: 8.0 +2023-03-21 03:12:08,170 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66957.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:12:10,601 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 03:12:11,299 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=66962.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:12:21,219 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 03:12:30,941 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 03:12:32,747 INFO [train.py:901] (1/2) Epoch 24, batch 2050, loss[loss=0.1536, simple_loss=0.2303, pruned_loss=0.03847, over 7288.00 frames. ], tot_loss[loss=0.1435, simple_loss=0.2227, pruned_loss=0.03212, over 1440215.42 frames. ], batch size: 49, lr: 6.85e-03, grad_scale: 8.0 +2023-03-21 03:12:40,295 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67018.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 03:12:45,161 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.240e+02 1.848e+02 2.104e+02 2.679e+02 4.101e+02, threshold=4.209e+02, percent-clipped=1.0 +2023-03-21 03:12:48,584 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 +2023-03-21 03:12:48,847 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3693, 1.2475, 1.4841, 1.8329, 1.5530, 1.7389, 1.3004, 1.7566], + device='cuda:1'), covar=tensor([0.1415, 0.2765, 0.1399, 0.1013, 0.1699, 0.1714, 0.1345, 0.1561], + device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0062, 0.0052, 0.0046, 0.0049, 0.0049, 0.0075, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:12:49,844 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4476, 3.1742, 2.2188, 3.6954, 2.7596, 3.4162, 1.6503, 2.1924], + device='cuda:1'), covar=tensor([0.0455, 0.0841, 0.2259, 0.0684, 0.0494, 0.0633, 0.3185, 0.1730], + device='cuda:1'), in_proj_covar=tensor([0.0268, 0.0249, 0.0284, 0.0260, 0.0265, 0.0257, 0.0248, 0.0269], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:12:58,206 INFO [train.py:901] (1/2) Epoch 24, batch 2100, loss[loss=0.1669, simple_loss=0.2444, pruned_loss=0.04467, over 7319.00 frames. ], tot_loss[loss=0.1434, simple_loss=0.2226, pruned_loss=0.03209, over 1440813.17 frames. ], batch size: 59, lr: 6.85e-03, grad_scale: 8.0 +2023-03-21 03:13:03,646 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 03:13:06,069 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 03:13:13,817 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67083.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:13:15,777 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67087.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 03:13:23,623 INFO [train.py:901] (1/2) Epoch 24, batch 2150, loss[loss=0.1468, simple_loss=0.2, pruned_loss=0.04679, over 6128.00 frames. ], tot_loss[loss=0.144, simple_loss=0.2231, pruned_loss=0.0325, over 1438761.01 frames. ], batch size: 26, lr: 6.85e-03, grad_scale: 8.0 +2023-03-21 03:13:26,235 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0801, 3.8899, 4.1896, 4.1048, 4.0764, 4.1773, 4.2204, 4.1270], + device='cuda:1'), covar=tensor([0.0036, 0.0079, 0.0032, 0.0034, 0.0033, 0.0029, 0.0026, 0.0048], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0059, 0.0052, 0.0049, 0.0048, 0.0053, 0.0046, 0.0065], + device='cuda:1'), out_proj_covar=tensor([8.1832e-05, 1.3419e-04, 1.0797e-04, 9.6688e-05, 9.3169e-05, 1.0272e-04, + 9.8984e-05, 1.3266e-04], device='cuda:1') +2023-03-21 03:13:36,124 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.281e+02 1.954e+02 2.289e+02 2.688e+02 5.693e+02, threshold=4.578e+02, percent-clipped=4.0 +2023-03-21 03:13:44,944 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67144.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:13:49,319 INFO [train.py:901] (1/2) Epoch 24, batch 2200, loss[loss=0.1435, simple_loss=0.2297, pruned_loss=0.02865, over 7246.00 frames. ], tot_loss[loss=0.144, simple_loss=0.2232, pruned_loss=0.03236, over 1438158.19 frames. ], batch size: 89, lr: 6.85e-03, grad_scale: 8.0 +2023-03-21 03:13:51,345 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 03:14:02,176 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67177.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 03:14:02,221 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9695, 2.4721, 1.9291, 2.4006, 2.7312, 2.8638, 2.2945, 2.1271], + device='cuda:1'), covar=tensor([0.2288, 0.0960, 0.3561, 0.0535, 0.0231, 0.0201, 0.0289, 0.0354], + device='cuda:1'), in_proj_covar=tensor([0.0264, 0.0233, 0.0266, 0.0267, 0.0178, 0.0175, 0.0204, 0.0219], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:14:05,193 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4867, 1.1134, 1.5271, 1.9036, 1.5704, 1.7612, 1.4527, 1.8002], + device='cuda:1'), covar=tensor([0.1816, 0.4102, 0.1471, 0.1121, 0.1635, 0.2123, 0.1996, 0.1999], + device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0063, 0.0052, 0.0047, 0.0049, 0.0049, 0.0076, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:14:15,373 INFO [train.py:901] (1/2) Epoch 24, batch 2250, loss[loss=0.139, simple_loss=0.2185, pruned_loss=0.02977, over 7335.00 frames. ], tot_loss[loss=0.144, simple_loss=0.2231, pruned_loss=0.03243, over 1438956.40 frames. ], batch size: 51, lr: 6.84e-03, grad_scale: 8.0 +2023-03-21 03:14:17,307 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 +2023-03-21 03:14:26,014 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 03:14:26,026 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 03:14:27,120 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=67225.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:14:28,517 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.343e+02 1.916e+02 2.226e+02 2.655e+02 7.651e+02, threshold=4.451e+02, percent-clipped=1.0 +2023-03-21 03:14:38,505 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 03:14:41,585 INFO [train.py:901] (1/2) Epoch 24, batch 2300, loss[loss=0.1468, simple_loss=0.2264, pruned_loss=0.03363, over 7350.00 frames. ], tot_loss[loss=0.1437, simple_loss=0.2226, pruned_loss=0.03241, over 1436913.05 frames. ], batch size: 73, lr: 6.84e-03, grad_scale: 8.0 +2023-03-21 03:14:51,428 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 +2023-03-21 03:15:06,586 INFO [train.py:901] (1/2) Epoch 24, batch 2350, loss[loss=0.1452, simple_loss=0.2325, pruned_loss=0.02891, over 7294.00 frames. ], tot_loss[loss=0.1442, simple_loss=0.2231, pruned_loss=0.03267, over 1440249.92 frames. ], batch size: 66, lr: 6.84e-03, grad_scale: 8.0 +2023-03-21 03:15:12,295 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67313.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 03:15:13,581 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.83 vs. limit=2.0 +2023-03-21 03:15:19,734 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.340e+02 1.854e+02 2.186e+02 2.625e+02 5.810e+02, threshold=4.373e+02, percent-clipped=2.0 +2023-03-21 03:15:25,803 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 03:15:32,504 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 03:15:32,976 INFO [train.py:901] (1/2) Epoch 24, batch 2400, loss[loss=0.1442, simple_loss=0.2346, pruned_loss=0.02692, over 7360.00 frames. ], tot_loss[loss=0.1438, simple_loss=0.223, pruned_loss=0.03227, over 1442029.90 frames. ], batch size: 73, lr: 6.84e-03, grad_scale: 8.0 +2023-03-21 03:15:40,851 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5564, 5.0428, 5.1060, 5.0736, 4.8479, 4.5665, 5.1464, 4.9615], + device='cuda:1'), covar=tensor([0.0407, 0.0357, 0.0378, 0.0424, 0.0319, 0.0366, 0.0317, 0.0429], + device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0237, 0.0182, 0.0183, 0.0143, 0.0213, 0.0185, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 03:15:43,241 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 03:15:45,761 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 03:15:50,358 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67387.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 03:15:59,087 INFO [train.py:901] (1/2) Epoch 24, batch 2450, loss[loss=0.1789, simple_loss=0.2567, pruned_loss=0.05052, over 6734.00 frames. ], tot_loss[loss=0.1432, simple_loss=0.2225, pruned_loss=0.03197, over 1440598.29 frames. ], batch size: 106, lr: 6.83e-03, grad_scale: 8.0 +2023-03-21 03:16:12,256 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.300e+02 1.866e+02 2.316e+02 2.804e+02 6.661e+02, threshold=4.633e+02, percent-clipped=5.0 +2023-03-21 03:16:13,322 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 03:16:15,822 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=67435.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:16:17,750 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67439.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:16:23,165 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-21 03:16:24,683 INFO [train.py:901] (1/2) Epoch 24, batch 2500, loss[loss=0.1772, simple_loss=0.2532, pruned_loss=0.05059, over 7342.00 frames. ], tot_loss[loss=0.1437, simple_loss=0.223, pruned_loss=0.03225, over 1440984.64 frames. ], batch size: 54, lr: 6.83e-03, grad_scale: 8.0 +2023-03-21 03:16:37,626 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 03:16:50,202 INFO [train.py:901] (1/2) Epoch 24, batch 2550, loss[loss=0.1429, simple_loss=0.2259, pruned_loss=0.02996, over 7315.00 frames. ], tot_loss[loss=0.1436, simple_loss=0.2231, pruned_loss=0.03208, over 1441546.41 frames. ], batch size: 59, lr: 6.83e-03, grad_scale: 8.0 +2023-03-21 03:16:55,919 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67513.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:17:03,298 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.348e+02 1.939e+02 2.232e+02 2.615e+02 5.776e+02, threshold=4.464e+02, percent-clipped=2.0 +2023-03-21 03:17:04,475 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9580, 3.1301, 3.7904, 3.7271, 3.7475, 3.8576, 3.7380, 3.7459], + device='cuda:1'), covar=tensor([0.0027, 0.0112, 0.0031, 0.0040, 0.0036, 0.0029, 0.0045, 0.0047], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0060, 0.0052, 0.0050, 0.0049, 0.0053, 0.0046, 0.0066], + device='cuda:1'), out_proj_covar=tensor([8.2156e-05, 1.3655e-04, 1.0864e-04, 9.8255e-05, 9.4769e-05, 1.0384e-04, + 9.9112e-05, 1.3397e-04], device='cuda:1') +2023-03-21 03:17:15,864 INFO [train.py:901] (1/2) Epoch 24, batch 2600, loss[loss=0.1362, simple_loss=0.2143, pruned_loss=0.02899, over 7274.00 frames. ], tot_loss[loss=0.1436, simple_loss=0.2228, pruned_loss=0.03219, over 1442366.72 frames. ], batch size: 77, lr: 6.83e-03, grad_scale: 8.0 +2023-03-21 03:17:15,932 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.9438, 4.4459, 4.5363, 4.9742, 4.9136, 4.9020, 4.4909, 4.5221], + device='cuda:1'), covar=tensor([0.0648, 0.2372, 0.1761, 0.0895, 0.0663, 0.0952, 0.0615, 0.0889], + device='cuda:1'), in_proj_covar=tensor([0.0220, 0.0341, 0.0263, 0.0260, 0.0195, 0.0328, 0.0198, 0.0242], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:17:26,310 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67574.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 03:17:29,910 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-03-21 03:17:39,334 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0199, 2.6357, 1.9070, 2.5275, 2.4720, 2.6395, 2.3302, 1.9264], + device='cuda:1'), covar=tensor([0.1992, 0.0842, 0.3289, 0.0531, 0.0181, 0.0183, 0.0303, 0.0290], + device='cuda:1'), in_proj_covar=tensor([0.0265, 0.0234, 0.0266, 0.0265, 0.0178, 0.0176, 0.0205, 0.0218], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:17:40,612 INFO [train.py:901] (1/2) Epoch 24, batch 2650, loss[loss=0.1509, simple_loss=0.2279, pruned_loss=0.03692, over 7345.00 frames. ], tot_loss[loss=0.1443, simple_loss=0.2236, pruned_loss=0.03252, over 1444196.76 frames. ], batch size: 63, lr: 6.82e-03, grad_scale: 8.0 +2023-03-21 03:17:44,859 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 +2023-03-21 03:17:45,620 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67613.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:17:46,371 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-03-21 03:17:49,530 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2726, 4.7490, 4.8056, 4.7102, 4.7004, 4.3707, 4.8488, 4.7179], + device='cuda:1'), covar=tensor([0.0502, 0.0422, 0.0389, 0.0505, 0.0313, 0.0359, 0.0328, 0.0436], + device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0235, 0.0180, 0.0181, 0.0144, 0.0211, 0.0185, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 03:17:52,864 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.362e+02 1.855e+02 2.123e+02 2.589e+02 4.353e+02, threshold=4.245e+02, percent-clipped=1.0 +2023-03-21 03:17:57,847 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0595, 3.9801, 3.2894, 3.5499, 2.9990, 2.0596, 1.8114, 4.1503], + device='cuda:1'), covar=tensor([0.0039, 0.0046, 0.0099, 0.0057, 0.0123, 0.0508, 0.0555, 0.0037], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0080, 0.0100, 0.0085, 0.0115, 0.0125, 0.0121, 0.0092], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 03:18:04,015 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 +2023-03-21 03:18:05,654 INFO [train.py:901] (1/2) Epoch 24, batch 2700, loss[loss=0.163, simple_loss=0.2421, pruned_loss=0.04194, over 7143.00 frames. ], tot_loss[loss=0.1444, simple_loss=0.2238, pruned_loss=0.03253, over 1445433.72 frames. ], batch size: 98, lr: 6.82e-03, grad_scale: 8.0 +2023-03-21 03:18:09,619 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=67661.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:18:18,981 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67680.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:18:24,889 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67692.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:18:30,272 INFO [train.py:901] (1/2) Epoch 24, batch 2750, loss[loss=0.1466, simple_loss=0.2255, pruned_loss=0.03384, over 7324.00 frames. ], tot_loss[loss=0.144, simple_loss=0.2232, pruned_loss=0.03242, over 1442658.37 frames. ], batch size: 75, lr: 6.82e-03, grad_scale: 8.0 +2023-03-21 03:18:36,605 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2859, 4.2860, 3.8963, 3.8434, 3.3408, 2.3217, 1.9591, 4.3934], + device='cuda:1'), covar=tensor([0.0041, 0.0036, 0.0075, 0.0054, 0.0108, 0.0485, 0.0528, 0.0037], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0080, 0.0100, 0.0085, 0.0115, 0.0125, 0.0121, 0.0093], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 03:18:37,054 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6114, 4.1958, 4.1000, 4.6313, 4.4859, 4.6093, 4.1041, 4.2465], + device='cuda:1'), covar=tensor([0.0904, 0.2404, 0.2355, 0.1096, 0.0959, 0.1248, 0.0810, 0.1061], + device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0352, 0.0271, 0.0271, 0.0202, 0.0341, 0.0205, 0.0249], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:18:42,802 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.352e+02 1.781e+02 2.129e+02 2.566e+02 3.925e+02, threshold=4.258e+02, percent-clipped=0.0 +2023-03-21 03:18:48,281 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67739.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:18:49,281 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67741.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:18:54,792 INFO [train.py:901] (1/2) Epoch 24, batch 2800, loss[loss=0.1521, simple_loss=0.2214, pruned_loss=0.04138, over 7312.00 frames. ], tot_loss[loss=0.1449, simple_loss=0.2241, pruned_loss=0.03284, over 1443167.56 frames. ], batch size: 49, lr: 6.81e-03, grad_scale: 16.0 +2023-03-21 03:18:54,917 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67753.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 03:18:58,806 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9238, 2.5353, 2.9701, 2.9419, 2.9425, 2.7704, 2.3321, 3.0770], + device='cuda:1'), covar=tensor([0.1304, 0.0804, 0.1552, 0.1255, 0.0926, 0.1064, 0.2557, 0.1118], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0060, 0.0046, 0.0045, 0.0045, 0.0042, 0.0061, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:18:59,904 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 +2023-03-21 03:19:19,864 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 03:19:19,923 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 03:19:19,948 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 03:19:19,966 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 03:19:20,138 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 03:19:20,153 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 03:19:20,165 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 03:19:20,264 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 03:19:20,343 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 03:19:20,358 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 03:19:20,375 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 03:19:20,879 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 03:19:20,885 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 03:19:20,893 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 03:19:21,037 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 03:19:21,040 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 03:19:21,098 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 03:19:26,448 INFO [train.py:901] (1/2) Epoch 25, batch 0, loss[loss=0.1437, simple_loss=0.2252, pruned_loss=0.03112, over 7284.00 frames. ], tot_loss[loss=0.1437, simple_loss=0.2252, pruned_loss=0.03112, over 7284.00 frames. ], batch size: 66, lr: 6.68e-03, grad_scale: 16.0 +2023-03-21 03:19:26,448 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 03:19:32,217 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5011, 4.2287, 3.9499, 4.5585, 4.4431, 4.6870, 4.2221, 4.4118], + device='cuda:1'), covar=tensor([0.0818, 0.2120, 0.2027, 0.1335, 0.0809, 0.1035, 0.0602, 0.0751], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0356, 0.0273, 0.0275, 0.0203, 0.0344, 0.0207, 0.0252], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:19:35,175 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1125, 2.8018, 2.1096, 3.4550, 2.1852, 2.8506, 1.5985, 2.1122], + device='cuda:1'), covar=tensor([0.0445, 0.0697, 0.2587, 0.0532, 0.0465, 0.0619, 0.3679, 0.1822], + device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0248, 0.0289, 0.0260, 0.0268, 0.0258, 0.0247, 0.0271], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:19:46,857 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2565, 1.5017, 1.3715, 1.4219, 1.4638, 1.6086, 1.3682, 1.2459], + device='cuda:1'), covar=tensor([0.0168, 0.0101, 0.0162, 0.0101, 0.0076, 0.0091, 0.0101, 0.0114], + device='cuda:1'), in_proj_covar=tensor([0.0028, 0.0026, 0.0026, 0.0028, 0.0025, 0.0025, 0.0028, 0.0035], + device='cuda:1'), out_proj_covar=tensor([3.2567e-05, 2.9842e-05, 2.9985e-05, 3.0858e-05, 2.9073e-05, 2.8247e-05, + 3.1645e-05, 4.0120e-05], device='cuda:1') +2023-03-21 03:19:52,672 INFO [train.py:935] (1/2) Epoch 25, validation: loss=0.1652, simple_loss=0.2538, pruned_loss=0.03834, over 1622729.00 frames. +2023-03-21 03:19:52,673 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 03:19:57,701 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=67787.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:19:59,643 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 03:20:10,374 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 03:20:14,621 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 +2023-03-21 03:20:16,812 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 03:20:17,730 INFO [train.py:901] (1/2) Epoch 25, batch 50, loss[loss=0.1621, simple_loss=0.2453, pruned_loss=0.03942, over 7243.00 frames. ], tot_loss[loss=0.1452, simple_loss=0.2246, pruned_loss=0.03289, over 326405.05 frames. ], batch size: 93, lr: 6.68e-03, grad_scale: 16.0 +2023-03-21 03:20:18,204 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.338e+02 1.908e+02 2.288e+02 2.779e+02 7.073e+02, threshold=4.575e+02, percent-clipped=8.0 +2023-03-21 03:20:19,247 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 03:20:22,142 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 03:20:40,062 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67869.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 03:20:40,500 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 03:20:40,987 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 03:20:43,893 INFO [train.py:901] (1/2) Epoch 25, batch 100, loss[loss=0.1316, simple_loss=0.2116, pruned_loss=0.02576, over 7343.00 frames. ], tot_loss[loss=0.1446, simple_loss=0.2233, pruned_loss=0.03296, over 574015.44 frames. ], batch size: 63, lr: 6.67e-03, grad_scale: 16.0 +2023-03-21 03:21:08,880 INFO [train.py:901] (1/2) Epoch 25, batch 150, loss[loss=0.1448, simple_loss=0.2274, pruned_loss=0.03107, over 7347.00 frames. ], tot_loss[loss=0.1434, simple_loss=0.2228, pruned_loss=0.03204, over 767748.99 frames. ], batch size: 63, lr: 6.67e-03, grad_scale: 16.0 +2023-03-21 03:21:09,366 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.311e+02 1.796e+02 2.230e+02 2.756e+02 5.619e+02, threshold=4.460e+02, percent-clipped=1.0 +2023-03-21 03:21:35,038 INFO [train.py:901] (1/2) Epoch 25, batch 200, loss[loss=0.1498, simple_loss=0.2312, pruned_loss=0.03426, over 7289.00 frames. ], tot_loss[loss=0.1426, simple_loss=0.2222, pruned_loss=0.03153, over 915686.00 frames. ], batch size: 57, lr: 6.67e-03, grad_scale: 16.0 +2023-03-21 03:21:37,229 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67981.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:21:39,187 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8379, 4.1020, 3.7765, 4.0955, 3.6022, 4.0861, 4.3249, 4.4190], + device='cuda:1'), covar=tensor([0.0237, 0.0155, 0.0203, 0.0154, 0.0382, 0.0290, 0.0228, 0.0157], + device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0118, 0.0108, 0.0112, 0.0106, 0.0095, 0.0095, 0.0091], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:21:41,630 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 03:21:46,145 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 03:21:56,618 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 03:22:04,536 INFO [train.py:901] (1/2) Epoch 25, batch 250, loss[loss=0.1501, simple_loss=0.2195, pruned_loss=0.04041, over 7263.00 frames. ], tot_loss[loss=0.1435, simple_loss=0.2229, pruned_loss=0.03209, over 1032114.66 frames. ], batch size: 47, lr: 6.67e-03, grad_scale: 16.0 +2023-03-21 03:22:04,988 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.332e+02 1.830e+02 2.222e+02 2.557e+02 4.530e+02, threshold=4.443e+02, percent-clipped=1.0 +2023-03-21 03:22:09,646 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68036.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:22:10,556 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 03:22:12,611 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68042.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:22:15,521 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68048.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 03:22:25,261 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 +2023-03-21 03:22:28,493 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3588, 1.5570, 1.2768, 1.3630, 1.5117, 1.5015, 1.4372, 1.0798], + device='cuda:1'), covar=tensor([0.0137, 0.0135, 0.0258, 0.0115, 0.0119, 0.0083, 0.0095, 0.0157], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0026, 0.0026, 0.0028, 0.0026, 0.0025, 0.0028, 0.0036], + device='cuda:1'), out_proj_covar=tensor([3.2890e-05, 2.9889e-05, 3.0096e-05, 3.1327e-05, 2.9238e-05, 2.8369e-05, + 3.1753e-05, 4.0257e-05], device='cuda:1') +2023-03-21 03:22:29,508 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4006, 1.4859, 1.2593, 1.4786, 1.5415, 1.5455, 1.4162, 1.1015], + device='cuda:1'), covar=tensor([0.0111, 0.0115, 0.0189, 0.0122, 0.0081, 0.0075, 0.0232, 0.0123], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0026, 0.0026, 0.0028, 0.0026, 0.0025, 0.0028, 0.0036], + device='cuda:1'), out_proj_covar=tensor([3.2858e-05, 2.9866e-05, 3.0067e-05, 3.1305e-05, 2.9213e-05, 2.8365e-05, + 3.1712e-05, 4.0243e-05], device='cuda:1') +2023-03-21 03:22:29,839 INFO [train.py:901] (1/2) Epoch 25, batch 300, loss[loss=0.1363, simple_loss=0.2147, pruned_loss=0.02893, over 7278.00 frames. ], tot_loss[loss=0.1429, simple_loss=0.2224, pruned_loss=0.03169, over 1125821.14 frames. ], batch size: 70, lr: 6.66e-03, grad_scale: 16.0 +2023-03-21 03:22:29,857 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 03:22:34,983 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68087.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 03:22:39,363 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 03:22:55,967 INFO [train.py:901] (1/2) Epoch 25, batch 350, loss[loss=0.1243, simple_loss=0.2112, pruned_loss=0.0187, over 7275.00 frames. ], tot_loss[loss=0.1431, simple_loss=0.2226, pruned_loss=0.03178, over 1197155.96 frames. ], batch size: 77, lr: 6.66e-03, grad_scale: 16.0 +2023-03-21 03:22:56,442 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.041e+02 1.796e+02 2.023e+02 2.447e+02 4.467e+02, threshold=4.045e+02, percent-clipped=1.0 +2023-03-21 03:23:03,831 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 +2023-03-21 03:23:06,619 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68148.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 03:23:13,497 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 03:23:17,034 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68169.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:23:20,908 INFO [train.py:901] (1/2) Epoch 25, batch 400, loss[loss=0.1633, simple_loss=0.2353, pruned_loss=0.04569, over 7214.00 frames. ], tot_loss[loss=0.1434, simple_loss=0.2227, pruned_loss=0.03206, over 1249310.24 frames. ], batch size: 50, lr: 6.66e-03, grad_scale: 16.0 +2023-03-21 03:23:29,581 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 +2023-03-21 03:23:30,049 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 +2023-03-21 03:23:42,385 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=68217.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:23:47,346 INFO [train.py:901] (1/2) Epoch 25, batch 450, loss[loss=0.1627, simple_loss=0.2409, pruned_loss=0.04227, over 6705.00 frames. ], tot_loss[loss=0.1437, simple_loss=0.2231, pruned_loss=0.03213, over 1290583.29 frames. ], batch size: 106, lr: 6.66e-03, grad_scale: 16.0 +2023-03-21 03:23:47,817 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.300e+02 1.902e+02 2.293e+02 2.811e+02 5.974e+02, threshold=4.586e+02, percent-clipped=4.0 +2023-03-21 03:23:53,320 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 03:23:53,814 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 03:23:58,317 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1056, 3.8266, 3.8804, 3.7949, 3.7625, 3.7318, 4.0222, 3.6687], + device='cuda:1'), covar=tensor([0.0100, 0.0148, 0.0100, 0.0142, 0.0380, 0.0101, 0.0131, 0.0152], + device='cuda:1'), in_proj_covar=tensor([0.0084, 0.0091, 0.0090, 0.0079, 0.0159, 0.0099, 0.0094, 0.0098], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:24:09,293 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9345, 2.3170, 1.8612, 2.9300, 2.2708, 2.8620, 2.2775, 2.6912], + device='cuda:1'), covar=tensor([0.1984, 0.0950, 0.3382, 0.0476, 0.0153, 0.0220, 0.0265, 0.0380], + device='cuda:1'), in_proj_covar=tensor([0.0267, 0.0237, 0.0269, 0.0266, 0.0179, 0.0178, 0.0206, 0.0220], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:24:12,137 INFO [train.py:901] (1/2) Epoch 25, batch 500, loss[loss=0.16, simple_loss=0.2387, pruned_loss=0.04065, over 7259.00 frames. ], tot_loss[loss=0.1438, simple_loss=0.2231, pruned_loss=0.03224, over 1327023.24 frames. ], batch size: 55, lr: 6.65e-03, grad_scale: 16.0 +2023-03-21 03:24:13,953 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 +2023-03-21 03:24:16,329 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3554, 4.3203, 3.6677, 3.8315, 3.5026, 2.4366, 2.0422, 4.4350], + device='cuda:1'), covar=tensor([0.0046, 0.0059, 0.0108, 0.0063, 0.0100, 0.0470, 0.0562, 0.0042], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0079, 0.0101, 0.0086, 0.0115, 0.0124, 0.0120, 0.0092], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 03:24:27,218 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 03:24:28,720 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 03:24:29,231 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 03:24:31,743 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 03:24:35,746 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 03:24:38,182 INFO [train.py:901] (1/2) Epoch 25, batch 550, loss[loss=0.1523, simple_loss=0.2364, pruned_loss=0.03407, over 7315.00 frames. ], tot_loss[loss=0.1434, simple_loss=0.2225, pruned_loss=0.03215, over 1353788.02 frames. ], batch size: 83, lr: 6.65e-03, grad_scale: 16.0 +2023-03-21 03:24:38,649 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.401e+02 1.821e+02 2.169e+02 2.559e+02 5.307e+02, threshold=4.339e+02, percent-clipped=1.0 +2023-03-21 03:24:42,798 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68336.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:24:43,260 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68337.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:24:47,115 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 03:24:47,273 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1776, 2.9106, 2.0474, 3.5290, 2.4357, 2.8975, 1.5089, 1.9977], + device='cuda:1'), covar=tensor([0.0488, 0.0958, 0.2767, 0.0638, 0.0536, 0.0545, 0.3290, 0.1882], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0250, 0.0291, 0.0263, 0.0270, 0.0261, 0.0250, 0.0272], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:24:48,678 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68348.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:24:55,025 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 03:24:57,954 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 03:25:00,160 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5160, 3.0677, 2.5522, 2.8545, 2.8606, 2.5983, 2.8145, 2.6507], + device='cuda:1'), covar=tensor([0.1323, 0.0395, 0.1320, 0.1343, 0.1024, 0.0589, 0.1139, 0.0843], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0050, 0.0056, 0.0050, 0.0049, 0.0051, 0.0050, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:25:04,174 INFO [train.py:901] (1/2) Epoch 25, batch 600, loss[loss=0.1303, simple_loss=0.2041, pruned_loss=0.02827, over 7202.00 frames. ], tot_loss[loss=0.1431, simple_loss=0.2222, pruned_loss=0.03197, over 1374357.59 frames. ], batch size: 39, lr: 6.65e-03, grad_scale: 16.0 +2023-03-21 03:25:06,199 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 03:25:07,732 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=68384.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:25:13,710 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=68396.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:25:15,208 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68399.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:25:21,574 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2777, 1.4083, 1.3876, 1.4576, 1.5671, 1.4672, 1.5042, 1.1418], + device='cuda:1'), covar=tensor([0.0139, 0.0125, 0.0145, 0.0141, 0.0092, 0.0101, 0.0094, 0.0120], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0026, 0.0026, 0.0028, 0.0026, 0.0025, 0.0028, 0.0036], + device='cuda:1'), out_proj_covar=tensor([3.2949e-05, 2.9959e-05, 3.0033e-05, 3.1599e-05, 2.9806e-05, 2.8585e-05, + 3.1497e-05, 4.0262e-05], device='cuda:1') +2023-03-21 03:25:22,907 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 03:25:29,366 INFO [train.py:901] (1/2) Epoch 25, batch 650, loss[loss=0.1181, simple_loss=0.1888, pruned_loss=0.02373, over 7172.00 frames. ], tot_loss[loss=0.1438, simple_loss=0.2228, pruned_loss=0.03235, over 1389720.46 frames. ], batch size: 39, lr: 6.65e-03, grad_scale: 16.0 +2023-03-21 03:25:29,821 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.426e+02 1.836e+02 2.195e+02 2.668e+02 4.895e+02, threshold=4.391e+02, percent-clipped=2.0 +2023-03-21 03:25:31,786 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 03:25:37,351 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68443.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 03:25:38,439 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3783, 3.3932, 2.2233, 3.7950, 2.6284, 3.1199, 1.7318, 2.0693], + device='cuda:1'), covar=tensor([0.0418, 0.0720, 0.2921, 0.0479, 0.0496, 0.0722, 0.3544, 0.2317], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0249, 0.0289, 0.0263, 0.0269, 0.0261, 0.0248, 0.0272], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:25:46,394 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68460.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:25:48,266 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 03:25:55,636 INFO [train.py:901] (1/2) Epoch 25, batch 700, loss[loss=0.1771, simple_loss=0.2514, pruned_loss=0.05138, over 7361.00 frames. ], tot_loss[loss=0.1437, simple_loss=0.2227, pruned_loss=0.03229, over 1399038.42 frames. ], batch size: 63, lr: 6.64e-03, grad_scale: 16.0 +2023-03-21 03:25:57,628 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 03:26:05,521 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 +2023-03-21 03:26:20,648 INFO [train.py:901] (1/2) Epoch 25, batch 750, loss[loss=0.1634, simple_loss=0.2409, pruned_loss=0.04297, over 7269.00 frames. ], tot_loss[loss=0.1441, simple_loss=0.223, pruned_loss=0.0326, over 1408498.99 frames. ], batch size: 52, lr: 6.64e-03, grad_scale: 16.0 +2023-03-21 03:26:21,132 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.241e+02 1.945e+02 2.256e+02 2.748e+02 5.442e+02, threshold=4.512e+02, percent-clipped=1.0 +2023-03-21 03:26:21,167 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 03:26:22,148 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 03:26:30,760 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0954, 4.6796, 4.5271, 5.1313, 4.9537, 5.0786, 4.4607, 4.6773], + device='cuda:1'), covar=tensor([0.0711, 0.2136, 0.2060, 0.0944, 0.0704, 0.1082, 0.0822, 0.1029], + device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0349, 0.0272, 0.0272, 0.0199, 0.0332, 0.0204, 0.0248], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:26:36,923 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 03:26:37,706 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 +2023-03-21 03:26:40,997 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 03:26:45,398 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-03-21 03:26:46,920 INFO [train.py:901] (1/2) Epoch 25, batch 800, loss[loss=0.1427, simple_loss=0.2244, pruned_loss=0.03056, over 7272.00 frames. ], tot_loss[loss=0.1438, simple_loss=0.2228, pruned_loss=0.03235, over 1416999.78 frames. ], batch size: 77, lr: 6.64e-03, grad_scale: 16.0 +2023-03-21 03:26:46,922 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 03:26:47,579 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68578.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:26:47,963 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 03:26:55,586 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3381, 2.7207, 2.4177, 2.7184, 2.5994, 2.2190, 2.5379, 2.4698], + device='cuda:1'), covar=tensor([0.0845, 0.0528, 0.0822, 0.0832, 0.0565, 0.0952, 0.1136, 0.0969], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0049, 0.0056, 0.0050, 0.0048, 0.0051, 0.0050, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:26:58,441 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 03:27:12,626 INFO [train.py:901] (1/2) Epoch 25, batch 850, loss[loss=0.1459, simple_loss=0.2271, pruned_loss=0.03232, over 7301.00 frames. ], tot_loss[loss=0.1433, simple_loss=0.2225, pruned_loss=0.03203, over 1423917.67 frames. ], batch size: 59, lr: 6.64e-03, grad_scale: 16.0 +2023-03-21 03:27:13,099 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.204e+02 1.927e+02 2.218e+02 2.601e+02 4.733e+02, threshold=4.436e+02, percent-clipped=1.0 +2023-03-21 03:27:17,615 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 03:27:17,624 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 03:27:17,698 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68637.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:27:19,319 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68639.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:27:23,680 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 03:27:26,613 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 03:27:38,146 INFO [train.py:901] (1/2) Epoch 25, batch 900, loss[loss=0.1309, simple_loss=0.2147, pruned_loss=0.02358, over 7330.00 frames. ], tot_loss[loss=0.1429, simple_loss=0.2223, pruned_loss=0.03172, over 1430297.29 frames. ], batch size: 59, lr: 6.64e-03, grad_scale: 16.0 +2023-03-21 03:27:39,225 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6005, 1.8463, 2.1136, 2.1275, 2.0295, 1.6549, 1.5404, 1.5734], + device='cuda:1'), covar=tensor([0.0794, 0.0474, 0.0318, 0.0131, 0.0490, 0.0560, 0.0407, 0.0322], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0030, 0.0029, 0.0028, 0.0029, 0.0029, 0.0033, 0.0031], + device='cuda:1'), out_proj_covar=tensor([7.7979e-05, 7.8611e-05, 7.3386e-05, 7.2808e-05, 7.5209e-05, 7.3568e-05, + 8.0284e-05, 8.0866e-05], device='cuda:1') +2023-03-21 03:27:42,085 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:27:42,138 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:27:45,096 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0521, 4.5232, 4.5379, 4.9993, 4.9274, 5.0039, 4.3470, 4.6594], + device='cuda:1'), covar=tensor([0.0666, 0.2401, 0.2233, 0.0978, 0.0778, 0.1048, 0.0851, 0.0941], + device='cuda:1'), in_proj_covar=tensor([0.0227, 0.0350, 0.0272, 0.0271, 0.0201, 0.0332, 0.0204, 0.0248], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:27:51,545 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7057, 4.2556, 4.1580, 4.7025, 4.5623, 4.6845, 4.0236, 4.3197], + device='cuda:1'), covar=tensor([0.0790, 0.2624, 0.2691, 0.1003, 0.0908, 0.1116, 0.0857, 0.1051], + device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0353, 0.0274, 0.0273, 0.0202, 0.0334, 0.0205, 0.0250], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:28:04,024 INFO [train.py:901] (1/2) Epoch 25, batch 950, loss[loss=0.1453, simple_loss=0.2243, pruned_loss=0.03311, over 7293.00 frames. ], tot_loss[loss=0.1425, simple_loss=0.2218, pruned_loss=0.03161, over 1432065.68 frames. ], batch size: 57, lr: 6.63e-03, grad_scale: 16.0 +2023-03-21 03:28:04,487 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.356e+02 1.864e+02 2.097e+02 2.567e+02 3.634e+02, threshold=4.194e+02, percent-clipped=0.0 +2023-03-21 03:28:04,531 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 03:28:12,032 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68743.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 03:28:12,058 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8150, 2.4854, 3.0575, 2.9114, 3.0327, 2.7660, 2.2219, 2.9704], + device='cuda:1'), covar=tensor([0.2036, 0.0973, 0.1291, 0.1134, 0.0909, 0.0988, 0.3212, 0.1210], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0060, 0.0046, 0.0045, 0.0046, 0.0042, 0.0062, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:28:13,569 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68746.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:28:17,948 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68755.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:28:26,751 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 03:28:28,765 INFO [train.py:901] (1/2) Epoch 25, batch 1000, loss[loss=0.1367, simple_loss=0.2261, pruned_loss=0.02366, over 7287.00 frames. ], tot_loss[loss=0.1422, simple_loss=0.2218, pruned_loss=0.03135, over 1432951.10 frames. ], batch size: 86, lr: 6.63e-03, grad_scale: 16.0 +2023-03-21 03:28:35,771 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=68791.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 03:28:48,258 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 03:28:49,998 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-03-21 03:28:55,273 INFO [train.py:901] (1/2) Epoch 25, batch 1050, loss[loss=0.1029, simple_loss=0.1797, pruned_loss=0.01299, over 7036.00 frames. ], tot_loss[loss=0.1427, simple_loss=0.2218, pruned_loss=0.03177, over 1432175.82 frames. ], batch size: 35, lr: 6.63e-03, grad_scale: 16.0 +2023-03-21 03:28:55,736 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 1.902e+02 2.261e+02 2.596e+02 5.077e+02, threshold=4.521e+02, percent-clipped=1.0 +2023-03-21 03:29:02,875 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0865, 2.5527, 1.6844, 3.1707, 2.9241, 2.9139, 2.0153, 2.5917], + device='cuda:1'), covar=tensor([0.1730, 0.0758, 0.3577, 0.0483, 0.0209, 0.0176, 0.0244, 0.0392], + device='cuda:1'), in_proj_covar=tensor([0.0259, 0.0232, 0.0263, 0.0261, 0.0178, 0.0174, 0.0203, 0.0216], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:29:09,946 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 03:29:14,065 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 03:29:20,590 INFO [train.py:901] (1/2) Epoch 25, batch 1100, loss[loss=0.1761, simple_loss=0.2579, pruned_loss=0.04713, over 6818.00 frames. ], tot_loss[loss=0.1427, simple_loss=0.222, pruned_loss=0.03169, over 1435497.86 frames. ], batch size: 107, lr: 6.63e-03, grad_scale: 16.0 +2023-03-21 03:29:31,286 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0491, 3.4023, 2.9533, 3.2543, 3.2934, 2.6927, 3.1865, 3.1026], + device='cuda:1'), covar=tensor([0.0952, 0.1070, 0.1411, 0.1477, 0.1341, 0.0732, 0.1180, 0.1074], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0049, 0.0057, 0.0050, 0.0048, 0.0051, 0.0051, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:29:43,304 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 03:29:43,827 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 03:29:46,773 INFO [train.py:901] (1/2) Epoch 25, batch 1150, loss[loss=0.128, simple_loss=0.2104, pruned_loss=0.02274, over 7343.00 frames. ], tot_loss[loss=0.1418, simple_loss=0.2209, pruned_loss=0.03133, over 1435796.74 frames. ], batch size: 44, lr: 6.62e-03, grad_scale: 16.0 +2023-03-21 03:29:47,272 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.304e+02 1.925e+02 2.200e+02 2.576e+02 4.325e+02, threshold=4.400e+02, percent-clipped=0.0 +2023-03-21 03:29:50,308 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68934.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:29:56,115 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 03:29:56,626 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 03:29:57,701 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5687, 4.1319, 4.0731, 4.6067, 4.4449, 4.5471, 3.8365, 4.1464], + device='cuda:1'), covar=tensor([0.0835, 0.2512, 0.2144, 0.0896, 0.0742, 0.1035, 0.0852, 0.1024], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0351, 0.0271, 0.0271, 0.0200, 0.0333, 0.0205, 0.0248], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:29:58,778 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8786, 2.1427, 2.2921, 2.2707, 1.9454, 1.9126, 1.8410, 1.6811], + device='cuda:1'), covar=tensor([0.0600, 0.0499, 0.0216, 0.0207, 0.0767, 0.0494, 0.0351, 0.0504], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0030, 0.0029, 0.0029, 0.0029, 0.0029, 0.0033, 0.0032], + device='cuda:1'), out_proj_covar=tensor([7.8249e-05, 7.8946e-05, 7.3953e-05, 7.3781e-05, 7.6013e-05, 7.4326e-05, + 8.0932e-05, 8.2093e-05], device='cuda:1') +2023-03-21 03:30:00,774 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3622, 3.4906, 3.3571, 3.4819, 3.3273, 3.3460, 3.6645, 3.7301], + device='cuda:1'), covar=tensor([0.0288, 0.0206, 0.0212, 0.0212, 0.0358, 0.0763, 0.0285, 0.0222], + device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0117, 0.0106, 0.0111, 0.0104, 0.0097, 0.0093, 0.0092], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:30:12,250 INFO [train.py:901] (1/2) Epoch 25, batch 1200, loss[loss=0.1599, simple_loss=0.2384, pruned_loss=0.04065, over 7291.00 frames. ], tot_loss[loss=0.1423, simple_loss=0.2214, pruned_loss=0.03163, over 1436315.75 frames. ], batch size: 68, lr: 6.62e-03, grad_scale: 16.0 +2023-03-21 03:30:25,210 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1685, 3.4032, 2.9584, 3.5415, 3.4739, 3.1491, 3.2540, 2.9933], + device='cuda:1'), covar=tensor([0.1557, 0.0858, 0.1589, 0.1197, 0.0785, 0.0758, 0.1149, 0.2357], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0049, 0.0057, 0.0051, 0.0048, 0.0052, 0.0051, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:30:30,034 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 03:30:37,943 INFO [train.py:901] (1/2) Epoch 25, batch 1250, loss[loss=0.1351, simple_loss=0.2184, pruned_loss=0.02588, over 7276.00 frames. ], tot_loss[loss=0.1425, simple_loss=0.2216, pruned_loss=0.03176, over 1437495.71 frames. ], batch size: 77, lr: 6.62e-03, grad_scale: 16.0 +2023-03-21 03:30:38,396 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.226e+02 1.819e+02 2.207e+02 2.633e+02 5.960e+02, threshold=4.414e+02, percent-clipped=3.0 +2023-03-21 03:30:44,991 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69041.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:30:46,091 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69043.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:30:52,002 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69055.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:30:52,388 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 03:30:57,024 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 03:30:58,050 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 03:31:04,023 INFO [train.py:901] (1/2) Epoch 25, batch 1300, loss[loss=0.1074, simple_loss=0.1845, pruned_loss=0.01515, over 7025.00 frames. ], tot_loss[loss=0.1433, simple_loss=0.2224, pruned_loss=0.0321, over 1438318.81 frames. ], batch size: 35, lr: 6.62e-03, grad_scale: 16.0 +2023-03-21 03:31:17,168 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=69103.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:31:17,721 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69104.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:31:21,473 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 03:31:23,480 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 03:31:27,042 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 03:31:29,074 INFO [train.py:901] (1/2) Epoch 25, batch 1350, loss[loss=0.1569, simple_loss=0.2253, pruned_loss=0.04428, over 7290.00 frames. ], tot_loss[loss=0.143, simple_loss=0.2222, pruned_loss=0.03185, over 1439254.98 frames. ], batch size: 86, lr: 6.61e-03, grad_scale: 16.0 +2023-03-21 03:31:29,535 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.299e+02 1.957e+02 2.209e+02 2.682e+02 1.203e+03, threshold=4.418e+02, percent-clipped=5.0 +2023-03-21 03:31:36,571 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8306, 1.9446, 2.1820, 2.0437, 1.9741, 2.1065, 1.7168, 1.5203], + device='cuda:1'), covar=tensor([0.0479, 0.0392, 0.0192, 0.0243, 0.0628, 0.0347, 0.0316, 0.0287], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0030, 0.0029, 0.0029, 0.0030, 0.0029, 0.0032, 0.0032], + device='cuda:1'), out_proj_covar=tensor([7.8275e-05, 7.8321e-05, 7.3814e-05, 7.3849e-05, 7.6265e-05, 7.4068e-05, + 8.0149e-05, 8.1991e-05], device='cuda:1') +2023-03-21 03:31:37,462 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 03:31:47,260 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69162.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:31:49,891 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4094, 1.1972, 1.4182, 1.8167, 1.4423, 1.7258, 1.2782, 1.8181], + device='cuda:1'), covar=tensor([0.2472, 0.4038, 0.1624, 0.0962, 0.2605, 0.1181, 0.1836, 0.2061], + device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0064, 0.0051, 0.0046, 0.0050, 0.0049, 0.0078, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:31:50,915 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69168.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:31:51,348 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6621, 5.1903, 5.2114, 5.1488, 5.0044, 4.6102, 5.2726, 5.0733], + device='cuda:1'), covar=tensor([0.0454, 0.0405, 0.0463, 0.0507, 0.0328, 0.0332, 0.0327, 0.0432], + device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0239, 0.0181, 0.0182, 0.0143, 0.0211, 0.0185, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 03:31:55,232 INFO [train.py:901] (1/2) Epoch 25, batch 1400, loss[loss=0.1392, simple_loss=0.2225, pruned_loss=0.02797, over 7343.00 frames. ], tot_loss[loss=0.1431, simple_loss=0.2223, pruned_loss=0.03198, over 1440515.52 frames. ], batch size: 73, lr: 6.61e-03, grad_scale: 16.0 +2023-03-21 03:32:10,689 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 03:32:18,759 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69223.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:32:19,765 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69225.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:32:20,618 INFO [train.py:901] (1/2) Epoch 25, batch 1450, loss[loss=0.1175, simple_loss=0.1945, pruned_loss=0.02022, over 7135.00 frames. ], tot_loss[loss=0.143, simple_loss=0.2221, pruned_loss=0.03198, over 1441504.22 frames. ], batch size: 41, lr: 6.61e-03, grad_scale: 16.0 +2023-03-21 03:32:21,090 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.249e+02 1.877e+02 2.365e+02 2.787e+02 5.081e+02, threshold=4.731e+02, percent-clipped=1.0 +2023-03-21 03:32:21,740 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69229.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:32:24,799 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69234.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:32:30,313 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69245.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:32:35,257 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 03:32:44,820 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69273.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:32:46,635 INFO [train.py:901] (1/2) Epoch 25, batch 1500, loss[loss=0.1331, simple_loss=0.213, pruned_loss=0.02662, over 7164.00 frames. ], tot_loss[loss=0.1431, simple_loss=0.2224, pruned_loss=0.03187, over 1441159.50 frames. ], batch size: 41, lr: 6.61e-03, grad_scale: 16.0 +2023-03-21 03:32:49,151 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=69282.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:32:51,202 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69286.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:32:51,560 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 03:33:01,160 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69306.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 03:33:08,148 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69320.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:33:12,045 INFO [train.py:901] (1/2) Epoch 25, batch 1550, loss[loss=0.132, simple_loss=0.2117, pruned_loss=0.02614, over 7284.00 frames. ], tot_loss[loss=0.1425, simple_loss=0.2222, pruned_loss=0.03145, over 1441574.52 frames. ], batch size: 77, lr: 6.60e-03, grad_scale: 16.0 +2023-03-21 03:33:12,524 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.197e+02 1.813e+02 2.080e+02 2.419e+02 4.351e+02, threshold=4.160e+02, percent-clipped=0.0 +2023-03-21 03:33:14,519 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 03:33:15,619 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69334.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:33:19,682 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69341.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:33:30,294 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-03-21 03:33:33,197 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69368.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:33:37,587 INFO [train.py:901] (1/2) Epoch 25, batch 1600, loss[loss=0.1414, simple_loss=0.2151, pruned_loss=0.03381, over 7317.00 frames. ], tot_loss[loss=0.1429, simple_loss=0.2222, pruned_loss=0.03179, over 1440993.04 frames. ], batch size: 59, lr: 6.60e-03, grad_scale: 16.0 +2023-03-21 03:33:39,749 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69381.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:33:40,858 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 +2023-03-21 03:33:43,699 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=69389.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:33:44,658 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 03:33:45,184 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 03:33:46,265 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69394.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:33:48,644 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 03:33:48,720 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69399.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:33:52,587 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7827, 5.3087, 5.3753, 5.3111, 5.0547, 4.8145, 5.3610, 5.1410], + device='cuda:1'), covar=tensor([0.0478, 0.0394, 0.0369, 0.0450, 0.0369, 0.0377, 0.0392, 0.0454], + device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0239, 0.0182, 0.0182, 0.0145, 0.0213, 0.0190, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 03:33:59,203 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 03:34:03,627 INFO [train.py:901] (1/2) Epoch 25, batch 1650, loss[loss=0.1314, simple_loss=0.218, pruned_loss=0.02244, over 7342.00 frames. ], tot_loss[loss=0.1429, simple_loss=0.2222, pruned_loss=0.03183, over 1441666.60 frames. ], batch size: 75, lr: 6.60e-03, grad_scale: 16.0 +2023-03-21 03:34:03,645 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 03:34:04,762 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 1.921e+02 2.303e+02 2.853e+02 4.440e+02, threshold=4.606e+02, percent-clipped=2.0 +2023-03-21 03:34:05,446 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69429.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:34:12,304 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 03:34:18,468 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69455.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:34:18,987 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8480, 2.7020, 1.9889, 3.3119, 2.1078, 2.7362, 1.2875, 1.9523], + device='cuda:1'), covar=tensor([0.0536, 0.0996, 0.2732, 0.0792, 0.0499, 0.0588, 0.3577, 0.1953], + device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0252, 0.0292, 0.0267, 0.0272, 0.0263, 0.0252, 0.0271], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:34:28,757 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 03:34:29,256 INFO [train.py:901] (1/2) Epoch 25, batch 1700, loss[loss=0.1359, simple_loss=0.2087, pruned_loss=0.03152, over 7323.00 frames. ], tot_loss[loss=0.1431, simple_loss=0.2224, pruned_loss=0.03189, over 1439601.30 frames. ], batch size: 44, lr: 6.60e-03, grad_scale: 16.0 +2023-03-21 03:34:32,708 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 03:34:43,759 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 03:34:50,923 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69518.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:34:53,827 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69524.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:34:55,265 INFO [train.py:901] (1/2) Epoch 25, batch 1750, loss[loss=0.1249, simple_loss=0.2069, pruned_loss=0.0214, over 7290.00 frames. ], tot_loss[loss=0.1426, simple_loss=0.222, pruned_loss=0.03164, over 1440327.75 frames. ], batch size: 66, lr: 6.59e-03, grad_scale: 16.0 +2023-03-21 03:34:55,735 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.346e+02 1.951e+02 2.380e+02 2.655e+02 3.975e+02, threshold=4.760e+02, percent-clipped=0.0 +2023-03-21 03:34:56,398 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2942, 4.2148, 3.5605, 3.8035, 3.2102, 2.3745, 2.0977, 4.3425], + device='cuda:1'), covar=tensor([0.0042, 0.0035, 0.0089, 0.0051, 0.0121, 0.0448, 0.0522, 0.0036], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0080, 0.0102, 0.0086, 0.0115, 0.0123, 0.0120, 0.0093], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 03:34:59,404 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8687, 1.7526, 1.9798, 2.3392, 1.9484, 2.4824, 2.3431, 2.1940], + device='cuda:1'), covar=tensor([0.3916, 0.4246, 0.3643, 0.2299, 0.3703, 0.3906, 0.2155, 0.3924], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0066, 0.0053, 0.0047, 0.0051, 0.0050, 0.0080, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:35:08,781 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 03:35:09,761 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 03:35:17,229 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 +2023-03-21 03:35:17,463 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7709, 5.2979, 5.3017, 5.2160, 5.0644, 4.8207, 5.3279, 5.1174], + device='cuda:1'), covar=tensor([0.0422, 0.0342, 0.0359, 0.0433, 0.0307, 0.0311, 0.0303, 0.0417], + device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0235, 0.0180, 0.0181, 0.0143, 0.0210, 0.0188, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 03:35:20,419 INFO [train.py:901] (1/2) Epoch 25, batch 1800, loss[loss=0.1222, simple_loss=0.2057, pruned_loss=0.01937, over 7334.00 frames. ], tot_loss[loss=0.1418, simple_loss=0.2212, pruned_loss=0.03124, over 1438321.95 frames. ], batch size: 44, lr: 6.59e-03, grad_scale: 16.0 +2023-03-21 03:35:22,585 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69581.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:35:32,005 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 03:35:34,063 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69601.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 03:35:45,611 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6013, 3.5369, 2.4588, 4.0864, 3.1918, 3.5575, 1.8366, 2.1679], + device='cuda:1'), covar=tensor([0.0471, 0.0974, 0.2771, 0.0506, 0.0467, 0.0630, 0.3406, 0.2204], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0251, 0.0289, 0.0265, 0.0269, 0.0262, 0.0252, 0.0269], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:35:46,447 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 03:35:46,924 INFO [train.py:901] (1/2) Epoch 25, batch 1850, loss[loss=0.1699, simple_loss=0.2499, pruned_loss=0.04494, over 6728.00 frames. ], tot_loss[loss=0.1418, simple_loss=0.221, pruned_loss=0.0313, over 1435834.81 frames. ], batch size: 106, lr: 6.59e-03, grad_scale: 16.0 +2023-03-21 03:35:47,417 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.867e+02 2.216e+02 2.571e+02 4.186e+02, threshold=4.433e+02, percent-clipped=0.0 +2023-03-21 03:35:47,973 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69629.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 03:35:55,950 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 03:36:11,059 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69674.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:36:12,009 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69676.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:36:12,404 INFO [train.py:901] (1/2) Epoch 25, batch 1900, loss[loss=0.1449, simple_loss=0.2241, pruned_loss=0.03287, over 7315.00 frames. ], tot_loss[loss=0.1421, simple_loss=0.2213, pruned_loss=0.03142, over 1436419.22 frames. ], batch size: 61, lr: 6.59e-03, grad_scale: 16.0 +2023-03-21 03:36:12,431 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 03:36:24,028 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69699.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:36:30,055 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3854, 1.6625, 1.4140, 1.7626, 1.7109, 1.4156, 1.6020, 1.2576], + device='cuda:1'), covar=tensor([0.0125, 0.0089, 0.0256, 0.0124, 0.0089, 0.0117, 0.0135, 0.0140], + device='cuda:1'), in_proj_covar=tensor([0.0029, 0.0027, 0.0027, 0.0029, 0.0027, 0.0025, 0.0028, 0.0036], + device='cuda:1'), out_proj_covar=tensor([3.3695e-05, 3.0152e-05, 3.0962e-05, 3.2379e-05, 3.1116e-05, 2.8451e-05, + 3.1764e-05, 4.0781e-05], device='cuda:1') +2023-03-21 03:36:36,407 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69724.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:36:36,841 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 03:36:37,853 INFO [train.py:901] (1/2) Epoch 25, batch 1950, loss[loss=0.1438, simple_loss=0.2215, pruned_loss=0.0331, over 7279.00 frames. ], tot_loss[loss=0.1425, simple_loss=0.2218, pruned_loss=0.03157, over 1439667.05 frames. ], batch size: 70, lr: 6.59e-03, grad_scale: 16.0 +2023-03-21 03:36:38,849 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.292e+02 1.839e+02 2.249e+02 2.694e+02 4.374e+02, threshold=4.499e+02, percent-clipped=0.0 +2023-03-21 03:36:41,969 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69735.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:36:45,986 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6322, 2.4697, 2.2723, 3.8932, 1.7013, 3.6903, 1.4258, 3.2340], + device='cuda:1'), covar=tensor([0.0119, 0.1184, 0.1651, 0.0167, 0.3433, 0.0176, 0.1121, 0.0395], + device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0257, 0.0277, 0.0197, 0.0265, 0.0207, 0.0251, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:36:47,834 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 03:36:47,891 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=69747.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:36:49,440 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69750.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:36:52,846 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 03:36:53,855 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 03:37:04,174 INFO [train.py:901] (1/2) Epoch 25, batch 2000, loss[loss=0.1489, simple_loss=0.2277, pruned_loss=0.03498, over 7282.00 frames. ], tot_loss[loss=0.1421, simple_loss=0.2213, pruned_loss=0.0314, over 1440222.21 frames. ], batch size: 66, lr: 6.58e-03, grad_scale: 16.0 +2023-03-21 03:37:11,628 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 03:37:22,432 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 03:37:25,002 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69818.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:37:27,928 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69824.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:37:28,421 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4383, 4.9229, 4.9603, 4.8856, 4.8318, 4.4343, 5.0122, 4.8018], + device='cuda:1'), covar=tensor([0.0514, 0.0433, 0.0412, 0.0506, 0.0318, 0.0367, 0.0324, 0.0459], + device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0238, 0.0181, 0.0181, 0.0142, 0.0211, 0.0188, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 03:37:29,323 INFO [train.py:901] (1/2) Epoch 25, batch 2050, loss[loss=0.1457, simple_loss=0.2168, pruned_loss=0.03733, over 7365.00 frames. ], tot_loss[loss=0.1423, simple_loss=0.2215, pruned_loss=0.03154, over 1440970.45 frames. ], batch size: 51, lr: 6.58e-03, grad_scale: 16.0 +2023-03-21 03:37:29,845 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 03:37:30,799 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.280e+02 1.910e+02 2.230e+02 2.522e+02 4.297e+02, threshold=4.460e+02, percent-clipped=0.0 +2023-03-21 03:37:45,183 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8013, 3.1401, 3.6020, 3.8908, 3.8198, 3.8230, 3.7793, 3.6155], + device='cuda:1'), covar=tensor([0.0028, 0.0110, 0.0035, 0.0028, 0.0030, 0.0026, 0.0045, 0.0052], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0061, 0.0052, 0.0050, 0.0049, 0.0053, 0.0046, 0.0066], + device='cuda:1'), out_proj_covar=tensor([8.0788e-05, 1.3618e-04, 1.0592e-04, 9.6646e-05, 9.3832e-05, 1.0158e-04, + 9.8818e-05, 1.3299e-04], device='cuda:1') +2023-03-21 03:37:50,231 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=69866.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:37:53,199 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=69872.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:37:55,677 INFO [train.py:901] (1/2) Epoch 25, batch 2100, loss[loss=0.1414, simple_loss=0.2238, pruned_loss=0.02953, over 7355.00 frames. ], tot_loss[loss=0.1425, simple_loss=0.2217, pruned_loss=0.03169, over 1440117.51 frames. ], batch size: 61, lr: 6.58e-03, grad_scale: 8.0 +2023-03-21 03:37:57,779 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69881.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:38:03,132 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 03:38:06,196 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5151, 1.2103, 1.8270, 1.9797, 1.7151, 2.0460, 1.7452, 2.0944], + device='cuda:1'), covar=tensor([0.1767, 0.3646, 0.0789, 0.1119, 0.2257, 0.1872, 0.2624, 0.2117], + device='cuda:1'), in_proj_covar=tensor([0.0063, 0.0067, 0.0053, 0.0050, 0.0053, 0.0051, 0.0082, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:38:06,562 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 03:38:07,560 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69901.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:38:20,447 INFO [train.py:901] (1/2) Epoch 25, batch 2150, loss[loss=0.1066, simple_loss=0.1676, pruned_loss=0.02285, over 6009.00 frames. ], tot_loss[loss=0.1428, simple_loss=0.2219, pruned_loss=0.03181, over 1442017.41 frames. ], batch size: 25, lr: 6.58e-03, grad_scale: 8.0 +2023-03-21 03:38:21,545 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=69929.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:38:21,608 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69929.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:38:21,980 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.326e+02 1.939e+02 2.247e+02 2.652e+02 6.279e+02, threshold=4.495e+02, percent-clipped=1.0 +2023-03-21 03:38:33,005 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=69949.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 03:38:46,549 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69976.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:38:46,914 INFO [train.py:901] (1/2) Epoch 25, batch 2200, loss[loss=0.1727, simple_loss=0.246, pruned_loss=0.04966, over 6755.00 frames. ], tot_loss[loss=0.143, simple_loss=0.2221, pruned_loss=0.03195, over 1440573.75 frames. ], batch size: 106, lr: 6.57e-03, grad_scale: 8.0 +2023-03-21 03:38:46,989 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=69977.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:38:51,960 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 03:38:53,233 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.50 vs. limit=5.0 +2023-03-21 03:39:11,486 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:39:11,539 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:39:12,917 INFO [train.py:901] (1/2) Epoch 25, batch 2250, loss[loss=0.1392, simple_loss=0.2213, pruned_loss=0.02851, over 7314.00 frames. ], tot_loss[loss=0.143, simple_loss=0.2223, pruned_loss=0.03191, over 1440570.92 frames. ], batch size: 59, lr: 6.57e-03, grad_scale: 8.0 +2023-03-21 03:39:14,117 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4217, 1.1013, 1.5706, 1.8637, 1.6224, 1.8014, 1.4014, 1.8298], + device='cuda:1'), covar=tensor([0.2090, 0.4114, 0.1481, 0.1346, 0.1857, 0.1665, 0.1657, 0.3659], + device='cuda:1'), in_proj_covar=tensor([0.0064, 0.0067, 0.0054, 0.0050, 0.0053, 0.0052, 0.0083, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:39:14,426 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.235e+02 1.881e+02 2.155e+02 2.675e+02 4.570e+02, threshold=4.310e+02, percent-clipped=1.0 +2023-03-21 03:39:14,518 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70030.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 03:39:25,356 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70050.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:39:26,792 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 03:39:27,288 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 03:39:29,374 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0695, 2.7619, 3.1804, 2.6701, 3.1539, 2.8263, 2.4581, 3.0581], + device='cuda:1'), covar=tensor([0.1534, 0.0703, 0.1397, 0.2328, 0.0817, 0.1293, 0.2904, 0.1889], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0060, 0.0046, 0.0046, 0.0046, 0.0043, 0.0063, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:39:36,209 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=70072.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:39:38,657 INFO [train.py:901] (1/2) Epoch 25, batch 2300, loss[loss=0.1447, simple_loss=0.2152, pruned_loss=0.03711, over 7268.00 frames. ], tot_loss[loss=0.1432, simple_loss=0.2227, pruned_loss=0.03186, over 1442034.99 frames. ], batch size: 47, lr: 6.57e-03, grad_scale: 8.0 +2023-03-21 03:39:39,149 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 03:39:49,217 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=70098.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:39:55,789 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70110.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:39:55,836 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8117, 2.7601, 2.7756, 2.8925, 2.6886, 2.5266, 3.0177, 2.2015], + device='cuda:1'), covar=tensor([0.0449, 0.0667, 0.0476, 0.0561, 0.0566, 0.0744, 0.0731, 0.1586], + device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0336, 0.0268, 0.0357, 0.0304, 0.0300, 0.0347, 0.0277], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:40:04,777 INFO [train.py:901] (1/2) Epoch 25, batch 2350, loss[loss=0.1294, simple_loss=0.2059, pruned_loss=0.02648, over 7150.00 frames. ], tot_loss[loss=0.1435, simple_loss=0.2229, pruned_loss=0.03203, over 1440646.06 frames. ], batch size: 41, lr: 6.57e-03, grad_scale: 8.0 +2023-03-21 03:40:06,252 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 1.829e+02 2.172e+02 2.649e+02 5.270e+02, threshold=4.344e+02, percent-clipped=1.0 +2023-03-21 03:40:18,083 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5830, 3.0118, 2.6855, 2.8004, 2.9124, 2.5082, 2.9154, 2.6228], + device='cuda:1'), covar=tensor([0.1521, 0.0663, 0.1425, 0.1428, 0.1035, 0.1660, 0.0789, 0.2060], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0049, 0.0058, 0.0051, 0.0049, 0.0052, 0.0050, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:40:26,011 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 03:40:27,120 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70171.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:40:29,682 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9234, 2.4090, 1.8373, 2.9922, 2.9120, 2.5206, 2.5010, 2.5743], + device='cuda:1'), covar=tensor([0.2085, 0.0967, 0.3418, 0.0753, 0.0220, 0.0186, 0.0265, 0.0302], + device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0235, 0.0264, 0.0265, 0.0178, 0.0179, 0.0206, 0.0219], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:40:29,992 INFO [train.py:901] (1/2) Epoch 25, batch 2400, loss[loss=0.1467, simple_loss=0.2304, pruned_loss=0.03144, over 7260.00 frames. ], tot_loss[loss=0.1436, simple_loss=0.2229, pruned_loss=0.03211, over 1441085.66 frames. ], batch size: 64, lr: 6.56e-03, grad_scale: 8.0 +2023-03-21 03:40:32,465 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 03:40:36,288 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-03-21 03:40:42,837 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 03:40:45,390 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 03:40:54,576 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 +2023-03-21 03:40:56,556 INFO [train.py:901] (1/2) Epoch 25, batch 2450, loss[loss=0.1643, simple_loss=0.2361, pruned_loss=0.04627, over 7264.00 frames. ], tot_loss[loss=0.1435, simple_loss=0.2227, pruned_loss=0.0321, over 1439496.07 frames. ], batch size: 52, lr: 6.56e-03, grad_scale: 8.0 +2023-03-21 03:40:58,016 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.945e+02 2.299e+02 2.753e+02 4.532e+02, threshold=4.598e+02, percent-clipped=1.0 +2023-03-21 03:41:00,649 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5330, 1.4993, 1.8666, 2.1930, 1.9295, 2.3331, 2.0646, 2.1827], + device='cuda:1'), covar=tensor([0.1673, 0.2694, 0.1185, 0.1361, 0.1716, 0.2515, 0.1830, 0.1757], + device='cuda:1'), in_proj_covar=tensor([0.0064, 0.0066, 0.0054, 0.0050, 0.0053, 0.0053, 0.0082, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:41:13,011 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70260.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:41:13,407 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 03:41:21,397 INFO [train.py:901] (1/2) Epoch 25, batch 2500, loss[loss=0.1365, simple_loss=0.2202, pruned_loss=0.02637, over 7323.00 frames. ], tot_loss[loss=0.143, simple_loss=0.2222, pruned_loss=0.03194, over 1441416.49 frames. ], batch size: 59, lr: 6.56e-03, grad_scale: 8.0 +2023-03-21 03:41:39,716 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 03:41:44,853 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70321.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:41:47,617 INFO [train.py:901] (1/2) Epoch 25, batch 2550, loss[loss=0.1487, simple_loss=0.2297, pruned_loss=0.0339, over 7276.00 frames. ], tot_loss[loss=0.1428, simple_loss=0.222, pruned_loss=0.03179, over 1442534.68 frames. ], batch size: 70, lr: 6.56e-03, grad_scale: 8.0 +2023-03-21 03:41:49,158 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.766e+02 2.062e+02 2.405e+02 6.552e+02, threshold=4.124e+02, percent-clipped=2.0 +2023-03-21 03:41:49,288 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70330.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 03:42:13,033 INFO [train.py:901] (1/2) Epoch 25, batch 2600, loss[loss=0.1803, simple_loss=0.2653, pruned_loss=0.04769, over 6694.00 frames. ], tot_loss[loss=0.143, simple_loss=0.222, pruned_loss=0.03197, over 1442066.41 frames. ], batch size: 106, lr: 6.56e-03, grad_scale: 8.0 +2023-03-21 03:42:13,579 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=70378.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 03:42:37,927 INFO [train.py:901] (1/2) Epoch 25, batch 2650, loss[loss=0.1346, simple_loss=0.2193, pruned_loss=0.02492, over 7265.00 frames. ], tot_loss[loss=0.1432, simple_loss=0.2222, pruned_loss=0.03207, over 1439913.47 frames. ], batch size: 89, lr: 6.55e-03, grad_scale: 8.0 +2023-03-21 03:42:39,419 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.240e+02 1.969e+02 2.211e+02 2.540e+02 3.994e+02, threshold=4.422e+02, percent-clipped=0.0 +2023-03-21 03:42:58,071 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70466.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:43:03,332 INFO [train.py:901] (1/2) Epoch 25, batch 2700, loss[loss=0.1596, simple_loss=0.2415, pruned_loss=0.03887, over 7359.00 frames. ], tot_loss[loss=0.1424, simple_loss=0.2213, pruned_loss=0.0317, over 1441246.23 frames. ], batch size: 63, lr: 6.55e-03, grad_scale: 8.0 +2023-03-21 03:43:13,157 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 +2023-03-21 03:43:28,019 INFO [train.py:901] (1/2) Epoch 25, batch 2750, loss[loss=0.1428, simple_loss=0.2207, pruned_loss=0.03244, over 7264.00 frames. ], tot_loss[loss=0.142, simple_loss=0.221, pruned_loss=0.0315, over 1439944.39 frames. ], batch size: 52, lr: 6.55e-03, grad_scale: 8.0 +2023-03-21 03:43:29,589 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.221e+02 1.852e+02 2.282e+02 2.643e+02 6.214e+02, threshold=4.563e+02, percent-clipped=1.0 +2023-03-21 03:43:43,919 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 +2023-03-21 03:43:52,390 INFO [train.py:901] (1/2) Epoch 25, batch 2800, loss[loss=0.1585, simple_loss=0.2419, pruned_loss=0.03754, over 6679.00 frames. ], tot_loss[loss=0.1427, simple_loss=0.2221, pruned_loss=0.0317, over 1439620.58 frames. ], batch size: 107, lr: 6.55e-03, grad_scale: 8.0 +2023-03-21 03:44:14,791 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 03:44:14,844 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 03:44:14,869 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 03:44:14,885 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 03:44:15,386 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 03:44:15,402 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 03:44:15,412 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 03:44:15,507 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 03:44:15,580 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 03:44:15,595 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 03:44:15,611 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 03:44:15,770 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 03:44:15,775 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 03:44:15,783 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 03:44:16,233 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 03:44:16,236 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 03:44:16,293 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 03:44:22,830 INFO [train.py:901] (1/2) Epoch 26, batch 0, loss[loss=0.1287, simple_loss=0.2085, pruned_loss=0.02442, over 7235.00 frames. ], tot_loss[loss=0.1287, simple_loss=0.2085, pruned_loss=0.02442, over 7235.00 frames. ], batch size: 45, lr: 6.42e-03, grad_scale: 8.0 +2023-03-21 03:44:22,830 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 03:44:34,540 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9396, 2.1340, 2.4173, 2.0086, 1.8993, 1.8261, 1.8612, 1.7925], + device='cuda:1'), covar=tensor([0.0582, 0.0338, 0.0104, 0.0208, 0.0495, 0.0579, 0.0189, 0.0288], + device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0029, 0.0030, 0.0028, 0.0029, 0.0029, 0.0032, 0.0032], + device='cuda:1'), out_proj_covar=tensor([7.7333e-05, 7.6813e-05, 7.4536e-05, 7.2799e-05, 7.5776e-05, 7.4223e-05, + 7.9874e-05, 8.1938e-05], device='cuda:1') +2023-03-21 03:44:39,461 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1818, 2.8479, 3.1769, 3.1113, 3.2630, 2.9543, 2.5576, 3.2175], + device='cuda:1'), covar=tensor([0.1170, 0.0568, 0.1175, 0.1328, 0.1003, 0.0806, 0.2727, 0.1178], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0057, 0.0044, 0.0044, 0.0044, 0.0041, 0.0060, 0.0045], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:44:42,359 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8531, 3.1776, 2.7951, 3.0297, 3.1547, 2.7613, 3.0792, 2.9538], + device='cuda:1'), covar=tensor([0.0733, 0.1012, 0.0778, 0.0955, 0.1008, 0.0898, 0.1060, 0.1573], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0050, 0.0059, 0.0052, 0.0050, 0.0053, 0.0051, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:44:43,451 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0336, 2.0705, 2.3075, 1.9411, 1.8839, 1.8031, 1.8686, 1.7316], + device='cuda:1'), covar=tensor([0.0294, 0.0428, 0.0181, 0.0238, 0.0382, 0.0431, 0.0282, 0.0323], + device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0029, 0.0030, 0.0028, 0.0029, 0.0029, 0.0032, 0.0032], + device='cuda:1'), out_proj_covar=tensor([7.7333e-05, 7.6813e-05, 7.4536e-05, 7.2799e-05, 7.5776e-05, 7.4223e-05, + 7.9874e-05, 8.1938e-05], device='cuda:1') +2023-03-21 03:44:48,532 INFO [train.py:935] (1/2) Epoch 26, validation: loss=0.166, simple_loss=0.2547, pruned_loss=0.03864, over 1622729.00 frames. +2023-03-21 03:44:48,532 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 03:44:56,106 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 03:44:56,174 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70616.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:45:00,923 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4529, 1.1106, 1.5194, 1.9274, 1.6472, 1.8436, 1.5042, 1.8900], + device='cuda:1'), covar=tensor([0.2156, 0.3634, 0.1102, 0.0832, 0.1123, 0.1727, 0.2013, 0.2043], + device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0066, 0.0053, 0.0049, 0.0053, 0.0052, 0.0082, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:45:03,814 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.284e+02 1.970e+02 2.312e+02 2.848e+02 5.222e+02, threshold=4.623e+02, percent-clipped=3.0 +2023-03-21 03:45:07,347 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 03:45:14,348 INFO [train.py:901] (1/2) Epoch 26, batch 50, loss[loss=0.1364, simple_loss=0.2191, pruned_loss=0.02683, over 7325.00 frames. ], tot_loss[loss=0.1421, simple_loss=0.2219, pruned_loss=0.03117, over 326765.65 frames. ], batch size: 61, lr: 6.42e-03, grad_scale: 8.0 +2023-03-21 03:45:14,350 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 03:45:16,191 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-21 03:45:16,349 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 03:45:19,496 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 03:45:36,336 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 03:45:36,812 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 03:45:39,748 INFO [train.py:901] (1/2) Epoch 26, batch 100, loss[loss=0.1301, simple_loss=0.2178, pruned_loss=0.02125, over 7243.00 frames. ], tot_loss[loss=0.1435, simple_loss=0.2232, pruned_loss=0.03194, over 573242.83 frames. ], batch size: 55, lr: 6.42e-03, grad_scale: 8.0 +2023-03-21 03:45:45,406 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7816, 2.9293, 3.6460, 3.6709, 3.8262, 3.8612, 3.7667, 3.7069], + device='cuda:1'), covar=tensor([0.0030, 0.0120, 0.0035, 0.0029, 0.0029, 0.0025, 0.0043, 0.0045], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0061, 0.0052, 0.0049, 0.0049, 0.0054, 0.0046, 0.0067], + device='cuda:1'), out_proj_covar=tensor([8.1877e-05, 1.3639e-04, 1.0578e-04, 9.4918e-05, 9.3391e-05, 1.0162e-04, + 9.8451e-05, 1.3327e-04], device='cuda:1') +2023-03-21 03:45:54,705 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.346e+02 1.824e+02 2.219e+02 2.666e+02 5.274e+02, threshold=4.438e+02, percent-clipped=1.0 +2023-03-21 03:46:05,898 INFO [train.py:901] (1/2) Epoch 26, batch 150, loss[loss=0.1043, simple_loss=0.1695, pruned_loss=0.01957, over 5961.00 frames. ], tot_loss[loss=0.1426, simple_loss=0.222, pruned_loss=0.03161, over 764531.67 frames. ], batch size: 25, lr: 6.41e-03, grad_scale: 8.0 +2023-03-21 03:46:13,528 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70766.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:46:14,541 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8052, 2.0060, 2.2849, 1.9941, 1.8683, 2.0084, 1.8859, 1.7551], + device='cuda:1'), covar=tensor([0.0473, 0.0484, 0.0220, 0.0161, 0.0601, 0.0432, 0.0252, 0.0250], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0030, 0.0030, 0.0029, 0.0030, 0.0029, 0.0033, 0.0033], + device='cuda:1'), out_proj_covar=tensor([7.8420e-05, 7.8202e-05, 7.6162e-05, 7.4366e-05, 7.7278e-05, 7.5074e-05, + 8.1475e-05, 8.3173e-05], device='cuda:1') +2023-03-21 03:46:31,892 INFO [train.py:901] (1/2) Epoch 26, batch 200, loss[loss=0.153, simple_loss=0.2354, pruned_loss=0.0353, over 7174.00 frames. ], tot_loss[loss=0.1416, simple_loss=0.2205, pruned_loss=0.03135, over 914211.88 frames. ], batch size: 98, lr: 6.41e-03, grad_scale: 8.0 +2023-03-21 03:46:37,437 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 03:46:38,519 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=70814.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:46:38,551 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0301, 4.5314, 4.5965, 4.5146, 4.5573, 4.0859, 4.6122, 4.4861], + device='cuda:1'), covar=tensor([0.0537, 0.0405, 0.0338, 0.0477, 0.0264, 0.0432, 0.0356, 0.0422], + device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0237, 0.0181, 0.0181, 0.0143, 0.0211, 0.0186, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 03:46:41,417 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 03:46:47,026 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.309e+02 1.835e+02 2.165e+02 2.609e+02 3.766e+02, threshold=4.330e+02, percent-clipped=0.0 +2023-03-21 03:46:48,080 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 03:46:57,517 INFO [train.py:901] (1/2) Epoch 26, batch 250, loss[loss=0.1389, simple_loss=0.2252, pruned_loss=0.02633, over 7327.00 frames. ], tot_loss[loss=0.1425, simple_loss=0.2216, pruned_loss=0.03171, over 1032478.50 frames. ], batch size: 61, lr: 6.41e-03, grad_scale: 8.0 +2023-03-21 03:47:00,583 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 03:47:21,665 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 03:47:23,105 INFO [train.py:901] (1/2) Epoch 26, batch 300, loss[loss=0.1701, simple_loss=0.2507, pruned_loss=0.04471, over 7315.00 frames. ], tot_loss[loss=0.143, simple_loss=0.222, pruned_loss=0.03197, over 1123903.01 frames. ], batch size: 59, lr: 6.41e-03, grad_scale: 8.0 +2023-03-21 03:47:30,693 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 03:47:30,822 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70916.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:47:38,304 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.193e+02 1.793e+02 2.181e+02 2.612e+02 5.300e+02, threshold=4.362e+02, percent-clipped=1.0 +2023-03-21 03:47:48,824 INFO [train.py:901] (1/2) Epoch 26, batch 350, loss[loss=0.1327, simple_loss=0.2134, pruned_loss=0.02599, over 7350.00 frames. ], tot_loss[loss=0.1422, simple_loss=0.2214, pruned_loss=0.03147, over 1195433.45 frames. ], batch size: 61, lr: 6.40e-03, grad_scale: 8.0 +2023-03-21 03:47:56,018 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=70964.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:47:57,881 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=2.02 vs. limit=2.0 +2023-03-21 03:48:05,457 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 03:48:10,054 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70992.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:48:14,659 INFO [train.py:901] (1/2) Epoch 26, batch 400, loss[loss=0.1449, simple_loss=0.2261, pruned_loss=0.03184, over 7272.00 frames. ], tot_loss[loss=0.1422, simple_loss=0.2215, pruned_loss=0.03139, over 1251270.59 frames. ], batch size: 64, lr: 6.40e-03, grad_scale: 8.0 +2023-03-21 03:48:29,729 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 1.976e+02 2.210e+02 2.576e+02 5.239e+02, threshold=4.420e+02, percent-clipped=2.0 +2023-03-21 03:48:29,881 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71030.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:48:40,691 INFO [train.py:901] (1/2) Epoch 26, batch 450, loss[loss=0.1462, simple_loss=0.2339, pruned_loss=0.02921, over 7234.00 frames. ], tot_loss[loss=0.1425, simple_loss=0.2219, pruned_loss=0.03156, over 1294477.19 frames. ], batch size: 93, lr: 6.40e-03, grad_scale: 8.0 +2023-03-21 03:48:41,879 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71053.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:48:46,191 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 03:48:46,206 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 03:48:46,783 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0247, 3.3027, 3.9004, 4.1725, 4.1190, 4.1036, 4.1506, 4.0618], + device='cuda:1'), covar=tensor([0.0036, 0.0121, 0.0038, 0.0028, 0.0030, 0.0032, 0.0029, 0.0046], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0061, 0.0052, 0.0049, 0.0049, 0.0053, 0.0046, 0.0067], + device='cuda:1'), out_proj_covar=tensor([8.1652e-05, 1.3744e-04, 1.0576e-04, 9.4878e-05, 9.1849e-05, 1.0030e-04, + 9.8006e-05, 1.3351e-04], device='cuda:1') +2023-03-21 03:48:57,367 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0254, 4.5309, 4.3933, 5.0626, 4.8784, 4.9886, 4.4483, 4.5878], + device='cuda:1'), covar=tensor([0.0710, 0.2249, 0.2074, 0.0869, 0.0741, 0.1094, 0.0655, 0.0976], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0355, 0.0277, 0.0274, 0.0202, 0.0341, 0.0209, 0.0249], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:49:01,663 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71091.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:49:05,703 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2858, 2.8331, 1.9467, 3.1466, 2.9985, 3.1217, 2.5945, 2.7895], + device='cuda:1'), covar=tensor([0.1905, 0.0756, 0.3751, 0.0625, 0.0186, 0.0155, 0.0341, 0.0312], + device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0237, 0.0266, 0.0266, 0.0180, 0.0182, 0.0208, 0.0218], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:49:06,535 INFO [train.py:901] (1/2) Epoch 26, batch 500, loss[loss=0.1467, simple_loss=0.2273, pruned_loss=0.0331, over 6652.00 frames. ], tot_loss[loss=0.1429, simple_loss=0.2222, pruned_loss=0.03178, over 1327438.79 frames. ], batch size: 106, lr: 6.40e-03, grad_scale: 8.0 +2023-03-21 03:49:19,615 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 03:49:21,079 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.306e+02 1.818e+02 2.239e+02 2.782e+02 4.854e+02, threshold=4.477e+02, percent-clipped=3.0 +2023-03-21 03:49:21,114 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 03:49:21,646 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 03:49:24,276 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 03:49:28,795 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 03:49:31,480 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5015, 1.0381, 1.6416, 1.7976, 1.5204, 1.8821, 1.4190, 1.7651], + device='cuda:1'), covar=tensor([0.1727, 0.3285, 0.1135, 0.0838, 0.2630, 0.1644, 0.1789, 0.2055], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0066, 0.0052, 0.0049, 0.0053, 0.0052, 0.0081, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:49:32,336 INFO [train.py:901] (1/2) Epoch 26, batch 550, loss[loss=0.1373, simple_loss=0.2158, pruned_loss=0.02939, over 7329.00 frames. ], tot_loss[loss=0.1424, simple_loss=0.2214, pruned_loss=0.03163, over 1351804.81 frames. ], batch size: 75, lr: 6.40e-03, grad_scale: 8.0 +2023-03-21 03:49:41,137 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 03:49:49,352 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 03:49:52,884 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 03:49:58,794 INFO [train.py:901] (1/2) Epoch 26, batch 600, loss[loss=0.1187, simple_loss=0.1965, pruned_loss=0.02046, over 7266.00 frames. ], tot_loss[loss=0.1414, simple_loss=0.2203, pruned_loss=0.03121, over 1369241.02 frames. ], batch size: 47, lr: 6.39e-03, grad_scale: 8.0 +2023-03-21 03:50:00,862 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 03:50:01,676 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.90 vs. limit=2.0 +2023-03-21 03:50:14,066 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.205e+02 1.854e+02 2.159e+02 2.545e+02 4.866e+02, threshold=4.318e+02, percent-clipped=2.0 +2023-03-21 03:50:17,075 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 03:50:24,640 INFO [train.py:901] (1/2) Epoch 26, batch 650, loss[loss=0.1315, simple_loss=0.2078, pruned_loss=0.02757, over 7279.00 frames. ], tot_loss[loss=0.1416, simple_loss=0.2208, pruned_loss=0.03122, over 1386988.02 frames. ], batch size: 47, lr: 6.39e-03, grad_scale: 8.0 +2023-03-21 03:50:25,653 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 03:50:43,962 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 03:50:50,430 INFO [train.py:901] (1/2) Epoch 26, batch 700, loss[loss=0.1573, simple_loss=0.2354, pruned_loss=0.03955, over 7349.00 frames. ], tot_loss[loss=0.1416, simple_loss=0.2208, pruned_loss=0.03124, over 1398077.67 frames. ], batch size: 61, lr: 6.39e-03, grad_scale: 8.0 +2023-03-21 03:50:52,486 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7828, 3.1266, 2.6500, 3.1021, 3.0406, 2.6717, 3.0220, 2.8374], + device='cuda:1'), covar=tensor([0.0665, 0.0600, 0.1731, 0.0726, 0.1113, 0.0854, 0.0892, 0.1013], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0052, 0.0060, 0.0054, 0.0052, 0.0055, 0.0053, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:50:53,468 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 03:51:04,333 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2143, 3.1303, 2.0985, 3.6192, 2.6489, 3.2830, 1.6419, 2.2125], + device='cuda:1'), covar=tensor([0.0412, 0.0962, 0.2613, 0.0587, 0.0562, 0.0566, 0.3621, 0.1781], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0247, 0.0291, 0.0264, 0.0268, 0.0262, 0.0248, 0.0266], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:51:05,617 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.388e+02 1.831e+02 2.185e+02 2.577e+02 5.485e+02, threshold=4.371e+02, percent-clipped=2.0 +2023-03-21 03:51:14,682 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71348.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:51:16,095 INFO [train.py:901] (1/2) Epoch 26, batch 750, loss[loss=0.1511, simple_loss=0.2315, pruned_loss=0.0354, over 7287.00 frames. ], tot_loss[loss=0.1414, simple_loss=0.2208, pruned_loss=0.03099, over 1410068.67 frames. ], batch size: 77, lr: 6.39e-03, grad_scale: 8.0 +2023-03-21 03:51:16,121 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 03:51:16,595 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 03:51:31,857 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 03:51:34,388 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71386.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:51:36,334 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 03:51:42,841 INFO [train.py:901] (1/2) Epoch 26, batch 800, loss[loss=0.1346, simple_loss=0.2244, pruned_loss=0.02237, over 7125.00 frames. ], tot_loss[loss=0.1405, simple_loss=0.22, pruned_loss=0.0305, over 1417326.07 frames. ], batch size: 98, lr: 6.38e-03, grad_scale: 8.0 +2023-03-21 03:51:42,862 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 03:51:44,382 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 03:51:44,985 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7935, 3.1056, 3.5691, 3.7986, 3.8149, 3.7818, 3.6666, 3.6223], + device='cuda:1'), covar=tensor([0.0028, 0.0107, 0.0038, 0.0029, 0.0028, 0.0027, 0.0039, 0.0045], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0060, 0.0051, 0.0049, 0.0048, 0.0052, 0.0045, 0.0065], + device='cuda:1'), out_proj_covar=tensor([8.0529e-05, 1.3461e-04, 1.0349e-04, 9.3362e-05, 9.0912e-05, 9.7539e-05, + 9.5822e-05, 1.2955e-04], device='cuda:1') +2023-03-21 03:51:55,028 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 03:51:57,554 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.309e+02 1.848e+02 2.128e+02 2.521e+02 4.378e+02, threshold=4.256e+02, percent-clipped=1.0 +2023-03-21 03:52:08,876 INFO [train.py:901] (1/2) Epoch 26, batch 850, loss[loss=0.1463, simple_loss=0.2234, pruned_loss=0.03466, over 7280.00 frames. ], tot_loss[loss=0.14, simple_loss=0.2193, pruned_loss=0.03038, over 1420703.13 frames. ], batch size: 52, lr: 6.38e-03, grad_scale: 8.0 +2023-03-21 03:52:10,231 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.90 vs. limit=2.0 +2023-03-21 03:52:14,502 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 03:52:14,507 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 03:52:20,085 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 03:52:24,054 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 03:52:34,600 INFO [train.py:901] (1/2) Epoch 26, batch 900, loss[loss=0.1501, simple_loss=0.2332, pruned_loss=0.03351, over 7114.00 frames. ], tot_loss[loss=0.1406, simple_loss=0.2197, pruned_loss=0.03079, over 1425722.81 frames. ], batch size: 98, lr: 6.38e-03, grad_scale: 8.0 +2023-03-21 03:52:38,732 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4277, 3.3204, 2.2156, 3.9099, 2.7883, 3.3011, 1.7599, 2.3232], + device='cuda:1'), covar=tensor([0.0384, 0.0873, 0.2527, 0.0463, 0.0549, 0.0485, 0.3041, 0.1855], + device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0247, 0.0291, 0.0264, 0.0268, 0.0263, 0.0247, 0.0266], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:52:49,100 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.291e+02 1.869e+02 2.202e+02 2.598e+02 5.253e+02, threshold=4.404e+02, percent-clipped=1.0 +2023-03-21 03:52:58,989 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5402, 3.3847, 2.6372, 3.9825, 3.0889, 3.3534, 2.0747, 2.6674], + device='cuda:1'), covar=tensor([0.0469, 0.0552, 0.2038, 0.0472, 0.0448, 0.0684, 0.2710, 0.1710], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0248, 0.0291, 0.0265, 0.0268, 0.0264, 0.0248, 0.0266], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:53:00,345 INFO [train.py:901] (1/2) Epoch 26, batch 950, loss[loss=0.1399, simple_loss=0.222, pruned_loss=0.02892, over 7317.00 frames. ], tot_loss[loss=0.1403, simple_loss=0.2192, pruned_loss=0.03073, over 1426539.23 frames. ], batch size: 83, lr: 6.38e-03, grad_scale: 8.0 +2023-03-21 03:53:02,396 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 03:53:24,426 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 03:53:26,267 INFO [train.py:901] (1/2) Epoch 26, batch 1000, loss[loss=0.1395, simple_loss=0.2201, pruned_loss=0.02946, over 7273.00 frames. ], tot_loss[loss=0.1408, simple_loss=0.22, pruned_loss=0.03082, over 1430627.71 frames. ], batch size: 52, lr: 6.38e-03, grad_scale: 8.0 +2023-03-21 03:53:32,311 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 +2023-03-21 03:53:34,039 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71616.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:53:41,505 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.311e+02 1.757e+02 2.158e+02 2.500e+02 4.041e+02, threshold=4.316e+02, percent-clipped=0.0 +2023-03-21 03:53:44,837 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-03-21 03:53:45,576 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 03:53:50,692 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71648.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:53:51,998 INFO [train.py:901] (1/2) Epoch 26, batch 1050, loss[loss=0.1628, simple_loss=0.2423, pruned_loss=0.04166, over 7232.00 frames. ], tot_loss[loss=0.1404, simple_loss=0.2198, pruned_loss=0.03056, over 1435277.89 frames. ], batch size: 93, lr: 6.37e-03, grad_scale: 8.0 +2023-03-21 03:54:05,936 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71677.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:54:08,260 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 03:54:10,393 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71686.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:54:12,787 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 03:54:15,291 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=71696.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:54:17,764 INFO [train.py:901] (1/2) Epoch 26, batch 1100, loss[loss=0.1419, simple_loss=0.2249, pruned_loss=0.0294, over 7338.00 frames. ], tot_loss[loss=0.1407, simple_loss=0.2197, pruned_loss=0.03079, over 1433868.92 frames. ], batch size: 73, lr: 6.37e-03, grad_scale: 8.0 +2023-03-21 03:54:17,880 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7435, 4.7939, 4.6811, 4.8204, 4.4872, 4.8110, 5.0481, 5.1060], + device='cuda:1'), covar=tensor([0.0161, 0.0127, 0.0121, 0.0121, 0.0240, 0.0165, 0.0183, 0.0144], + device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0119, 0.0107, 0.0113, 0.0106, 0.0096, 0.0096, 0.0091], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:54:25,126 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.8652, 4.3612, 4.3040, 4.8491, 4.7270, 4.8094, 4.3031, 4.4417], + device='cuda:1'), covar=tensor([0.0961, 0.2560, 0.2181, 0.1006, 0.0855, 0.1168, 0.0793, 0.0962], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0356, 0.0279, 0.0278, 0.0209, 0.0345, 0.0210, 0.0251], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:54:33,012 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.196e+02 1.836e+02 2.092e+02 2.603e+02 5.313e+02, threshold=4.183e+02, percent-clipped=2.0 +2023-03-21 03:54:34,207 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0762, 2.8621, 3.1489, 3.2260, 3.0986, 2.9306, 2.6480, 3.0015], + device='cuda:1'), covar=tensor([0.1324, 0.0796, 0.1425, 0.1002, 0.0938, 0.0951, 0.2217, 0.1930], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0061, 0.0046, 0.0045, 0.0045, 0.0042, 0.0062, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:54:35,102 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=71734.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:54:42,633 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 03:54:42,645 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 03:54:42,882 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-03-21 03:54:44,134 INFO [train.py:901] (1/2) Epoch 26, batch 1150, loss[loss=0.1445, simple_loss=0.2308, pruned_loss=0.02911, over 7282.00 frames. ], tot_loss[loss=0.1414, simple_loss=0.2204, pruned_loss=0.03122, over 1434827.31 frames. ], batch size: 77, lr: 6.37e-03, grad_scale: 8.0 +2023-03-21 03:54:54,747 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 03:54:55,241 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 03:55:10,332 INFO [train.py:901] (1/2) Epoch 26, batch 1200, loss[loss=0.1146, simple_loss=0.1926, pruned_loss=0.01828, over 7137.00 frames. ], tot_loss[loss=0.1421, simple_loss=0.2213, pruned_loss=0.03143, over 1438231.84 frames. ], batch size: 41, lr: 6.37e-03, grad_scale: 8.0 +2023-03-21 03:55:14,959 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9423, 2.6078, 3.0501, 2.9350, 2.8158, 2.9783, 2.4673, 2.8956], + device='cuda:1'), covar=tensor([0.1677, 0.0948, 0.1442, 0.1306, 0.1472, 0.0883, 0.2196, 0.2139], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0061, 0.0046, 0.0045, 0.0045, 0.0042, 0.0063, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:55:24,518 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5537, 4.0658, 3.9683, 4.4916, 4.3480, 4.5053, 4.0196, 4.0729], + device='cuda:1'), covar=tensor([0.0857, 0.2508, 0.2106, 0.1141, 0.0918, 0.1188, 0.0848, 0.1083], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0357, 0.0280, 0.0278, 0.0209, 0.0342, 0.0210, 0.0251], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:55:24,897 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.304e+02 1.827e+02 2.248e+02 2.679e+02 4.033e+02, threshold=4.496e+02, percent-clipped=0.0 +2023-03-21 03:55:27,470 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 03:55:35,985 INFO [train.py:901] (1/2) Epoch 26, batch 1250, loss[loss=0.1502, simple_loss=0.2319, pruned_loss=0.03422, over 7250.00 frames. ], tot_loss[loss=0.1428, simple_loss=0.2221, pruned_loss=0.03179, over 1436840.96 frames. ], batch size: 47, lr: 6.36e-03, grad_scale: 16.0 +2023-03-21 03:55:51,157 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 03:55:54,704 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 03:55:56,278 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 03:56:01,834 INFO [train.py:901] (1/2) Epoch 26, batch 1300, loss[loss=0.1342, simple_loss=0.2209, pruned_loss=0.02376, over 7317.00 frames. ], tot_loss[loss=0.1419, simple_loss=0.2213, pruned_loss=0.03128, over 1437175.25 frames. ], batch size: 59, lr: 6.36e-03, grad_scale: 16.0 +2023-03-21 03:56:07,308 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-03-21 03:56:08,531 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0810, 4.6441, 4.7459, 4.6715, 4.6784, 4.1817, 4.7616, 4.6237], + device='cuda:1'), covar=tensor([0.0601, 0.0468, 0.0439, 0.0587, 0.0331, 0.0437, 0.0347, 0.0508], + device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0238, 0.0183, 0.0182, 0.0145, 0.0211, 0.0184, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 03:56:11,664 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71919.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:56:13,276 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-03-21 03:56:13,886 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 +2023-03-21 03:56:17,020 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.296e+02 1.780e+02 2.070e+02 2.538e+02 4.671e+02, threshold=4.140e+02, percent-clipped=1.0 +2023-03-21 03:56:19,555 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 03:56:21,554 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 03:56:24,983 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 03:56:27,522 INFO [train.py:901] (1/2) Epoch 26, batch 1350, loss[loss=0.1352, simple_loss=0.2137, pruned_loss=0.02836, over 7316.00 frames. ], tot_loss[loss=0.1425, simple_loss=0.2218, pruned_loss=0.03156, over 1439848.57 frames. ], batch size: 59, lr: 6.36e-03, grad_scale: 16.0 +2023-03-21 03:56:29,125 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0121, 4.5748, 4.4288, 4.9114, 4.8281, 4.9643, 4.3774, 4.5817], + device='cuda:1'), covar=tensor([0.0671, 0.2239, 0.2245, 0.1124, 0.0814, 0.1062, 0.0750, 0.1018], + device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0358, 0.0281, 0.0279, 0.0210, 0.0346, 0.0210, 0.0254], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:56:35,644 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 03:56:38,796 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71972.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:56:42,893 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71980.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:56:50,919 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9891, 4.0302, 3.9369, 4.0592, 3.6574, 4.1676, 4.3316, 4.4065], + device='cuda:1'), covar=tensor([0.0184, 0.0169, 0.0176, 0.0150, 0.0380, 0.0214, 0.0255, 0.0180], + device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0116, 0.0105, 0.0112, 0.0104, 0.0093, 0.0094, 0.0089], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:56:51,453 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0285, 1.7464, 2.1151, 2.0781, 1.7253, 2.2246, 2.0236, 1.8521], + device='cuda:1'), covar=tensor([0.0567, 0.0556, 0.0404, 0.0192, 0.0681, 0.0354, 0.0277, 0.0347], + device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0031, 0.0031, 0.0030, 0.0031, 0.0031, 0.0034, 0.0033], + device='cuda:1'), out_proj_covar=tensor([8.0975e-05, 8.0405e-05, 7.8529e-05, 7.6527e-05, 7.8954e-05, 7.8220e-05, + 8.3880e-05, 8.4304e-05], device='cuda:1') +2023-03-21 03:56:57,012 INFO [train.py:901] (1/2) Epoch 26, batch 1400, loss[loss=0.1314, simple_loss=0.2249, pruned_loss=0.01898, over 7316.00 frames. ], tot_loss[loss=0.1424, simple_loss=0.2219, pruned_loss=0.0315, over 1440630.49 frames. ], batch size: 80, lr: 6.36e-03, grad_scale: 16.0 +2023-03-21 03:57:11,259 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 03:57:12,292 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 1.905e+02 2.148e+02 2.727e+02 3.705e+02, threshold=4.295e+02, percent-clipped=0.0 +2023-03-21 03:57:18,741 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 +2023-03-21 03:57:23,434 INFO [train.py:901] (1/2) Epoch 26, batch 1450, loss[loss=0.144, simple_loss=0.218, pruned_loss=0.03496, over 7221.00 frames. ], tot_loss[loss=0.1421, simple_loss=0.2219, pruned_loss=0.03121, over 1442460.84 frames. ], batch size: 45, lr: 6.36e-03, grad_scale: 16.0 +2023-03-21 03:57:37,029 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 03:57:49,177 INFO [train.py:901] (1/2) Epoch 26, batch 1500, loss[loss=0.1494, simple_loss=0.2325, pruned_loss=0.03316, over 7275.00 frames. ], tot_loss[loss=0.1413, simple_loss=0.2211, pruned_loss=0.03074, over 1440420.65 frames. ], batch size: 66, lr: 6.35e-03, grad_scale: 16.0 +2023-03-21 03:57:52,250 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 03:58:03,441 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72129.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:58:03,796 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.319e+02 1.929e+02 2.352e+02 2.805e+02 7.015e+02, threshold=4.703e+02, percent-clipped=5.0 +2023-03-21 03:58:15,069 INFO [train.py:901] (1/2) Epoch 26, batch 1550, loss[loss=0.1475, simple_loss=0.2315, pruned_loss=0.03174, over 7351.00 frames. ], tot_loss[loss=0.1409, simple_loss=0.2206, pruned_loss=0.03056, over 1439427.66 frames. ], batch size: 63, lr: 6.35e-03, grad_scale: 16.0 +2023-03-21 03:58:17,057 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 03:58:35,451 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72190.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:58:41,182 INFO [train.py:901] (1/2) Epoch 26, batch 1600, loss[loss=0.1366, simple_loss=0.2175, pruned_loss=0.02785, over 7249.00 frames. ], tot_loss[loss=0.1409, simple_loss=0.2209, pruned_loss=0.03045, over 1439658.38 frames. ], batch size: 55, lr: 6.35e-03, grad_scale: 16.0 +2023-03-21 03:58:47,728 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 03:58:48,728 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 03:58:52,201 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 03:58:56,363 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 1.815e+02 2.080e+02 2.642e+02 5.040e+02, threshold=4.161e+02, percent-clipped=1.0 +2023-03-21 03:59:02,443 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 03:59:06,439 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 03:59:06,851 INFO [train.py:901] (1/2) Epoch 26, batch 1650, loss[loss=0.1658, simple_loss=0.2419, pruned_loss=0.04483, over 7278.00 frames. ], tot_loss[loss=0.1412, simple_loss=0.2211, pruned_loss=0.03062, over 1442553.54 frames. ], batch size: 66, lr: 6.35e-03, grad_scale: 16.0 +2023-03-21 03:59:15,596 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 03:59:18,181 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72272.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:59:19,636 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72275.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:59:31,552 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 03:59:31,703 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9451, 2.9383, 1.9517, 3.3380, 2.5399, 3.0798, 1.4476, 1.9536], + device='cuda:1'), covar=tensor([0.0437, 0.0853, 0.2972, 0.0716, 0.0584, 0.0585, 0.3840, 0.1890], + device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0252, 0.0295, 0.0268, 0.0270, 0.0267, 0.0250, 0.0271], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 03:59:32,566 INFO [train.py:901] (1/2) Epoch 26, batch 1700, loss[loss=0.1071, simple_loss=0.1727, pruned_loss=0.02077, over 6449.00 frames. ], tot_loss[loss=0.1416, simple_loss=0.2214, pruned_loss=0.03091, over 1441103.99 frames. ], batch size: 28, lr: 6.34e-03, grad_scale: 16.0 +2023-03-21 03:59:35,123 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 03:59:42,994 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=72320.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:59:45,961 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 03:59:46,109 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72326.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 03:59:47,965 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.249e+02 1.865e+02 2.256e+02 2.622e+02 4.805e+02, threshold=4.513e+02, percent-clipped=4.0 +2023-03-21 03:59:51,628 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3900, 3.9161, 3.7903, 4.4789, 4.2234, 4.3180, 3.9267, 3.9077], + device='cuda:1'), covar=tensor([0.0942, 0.2648, 0.2610, 0.0997, 0.1061, 0.1274, 0.0844, 0.1393], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0359, 0.0280, 0.0277, 0.0208, 0.0342, 0.0210, 0.0253], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 03:59:59,235 INFO [train.py:901] (1/2) Epoch 26, batch 1750, loss[loss=0.1562, simple_loss=0.2339, pruned_loss=0.03924, over 7359.00 frames. ], tot_loss[loss=0.1414, simple_loss=0.2213, pruned_loss=0.0307, over 1440843.73 frames. ], batch size: 73, lr: 6.34e-03, grad_scale: 16.0 +2023-03-21 04:00:07,079 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 +2023-03-21 04:00:09,749 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 04:00:10,748 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 04:00:17,600 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72387.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:00:25,534 INFO [train.py:901] (1/2) Epoch 26, batch 1800, loss[loss=0.1335, simple_loss=0.2114, pruned_loss=0.02777, over 7339.00 frames. ], tot_loss[loss=0.1407, simple_loss=0.2206, pruned_loss=0.03039, over 1441647.41 frames. ], batch size: 54, lr: 6.34e-03, grad_scale: 16.0 +2023-03-21 04:00:25,978 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-21 04:00:33,766 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 04:00:35,349 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6449, 4.1534, 4.0478, 4.5798, 4.4672, 4.5349, 3.9656, 4.1249], + device='cuda:1'), covar=tensor([0.0935, 0.2671, 0.2441, 0.1212, 0.0913, 0.1334, 0.1054, 0.1356], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0361, 0.0281, 0.0280, 0.0208, 0.0344, 0.0211, 0.0256], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:00:40,213 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.401e+02 1.894e+02 2.228e+02 2.651e+02 4.777e+02, threshold=4.456e+02, percent-clipped=1.0 +2023-03-21 04:00:43,987 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9186, 3.7954, 3.7881, 3.6976, 3.6243, 3.5804, 3.9102, 3.3887], + device='cuda:1'), covar=tensor([0.0179, 0.0148, 0.0131, 0.0190, 0.0491, 0.0140, 0.0174, 0.0230], + device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0092, 0.0091, 0.0082, 0.0160, 0.0102, 0.0096, 0.0102], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:00:47,435 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 04:00:51,404 INFO [train.py:901] (1/2) Epoch 26, batch 1850, loss[loss=0.1544, simple_loss=0.2384, pruned_loss=0.03518, over 7131.00 frames. ], tot_loss[loss=0.1417, simple_loss=0.2216, pruned_loss=0.03092, over 1439936.57 frames. ], batch size: 98, lr: 6.34e-03, grad_scale: 16.0 +2023-03-21 04:00:56,900 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 04:01:09,230 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72485.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:01:14,197 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 04:01:17,125 INFO [train.py:901] (1/2) Epoch 26, batch 1900, loss[loss=0.1276, simple_loss=0.2019, pruned_loss=0.02664, over 7229.00 frames. ], tot_loss[loss=0.1412, simple_loss=0.2211, pruned_loss=0.03063, over 1442039.65 frames. ], batch size: 45, lr: 6.34e-03, grad_scale: 16.0 +2023-03-21 04:01:32,356 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.186e+02 1.886e+02 2.210e+02 2.551e+02 4.752e+02, threshold=4.419e+02, percent-clipped=1.0 +2023-03-21 04:01:33,759 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 +2023-03-21 04:01:40,014 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 04:01:42,962 INFO [train.py:901] (1/2) Epoch 26, batch 1950, loss[loss=0.1643, simple_loss=0.2353, pruned_loss=0.0466, over 7210.00 frames. ], tot_loss[loss=0.1408, simple_loss=0.2205, pruned_loss=0.0305, over 1439564.90 frames. ], batch size: 50, lr: 6.33e-03, grad_scale: 16.0 +2023-03-21 04:01:49,411 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.09 vs. limit=2.0 +2023-03-21 04:01:51,035 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 04:01:55,816 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5917, 4.0846, 3.9493, 4.6385, 4.4479, 4.4949, 4.0313, 4.0877], + device='cuda:1'), covar=tensor([0.0866, 0.2598, 0.2637, 0.0931, 0.0887, 0.1209, 0.0749, 0.1127], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0360, 0.0277, 0.0277, 0.0205, 0.0341, 0.0208, 0.0254], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:01:55,852 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72575.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:01:56,256 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 04:01:56,768 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 04:02:02,043 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6870, 3.2602, 3.5458, 3.4761, 3.1590, 2.9482, 3.7323, 2.7594], + device='cuda:1'), covar=tensor([0.0398, 0.0425, 0.0446, 0.0533, 0.0686, 0.0763, 0.0570, 0.1470], + device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0337, 0.0269, 0.0362, 0.0305, 0.0298, 0.0343, 0.0276], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:02:09,051 INFO [train.py:901] (1/2) Epoch 26, batch 2000, loss[loss=0.1361, simple_loss=0.2146, pruned_loss=0.02878, over 7249.00 frames. ], tot_loss[loss=0.141, simple_loss=0.2206, pruned_loss=0.03068, over 1440769.64 frames. ], batch size: 47, lr: 6.33e-03, grad_scale: 16.0 +2023-03-21 04:02:09,161 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1212, 4.5655, 4.6691, 4.6046, 4.5937, 4.1678, 4.6936, 4.5519], + device='cuda:1'), covar=tensor([0.0475, 0.0442, 0.0351, 0.0470, 0.0303, 0.0422, 0.0306, 0.0457], + device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0236, 0.0179, 0.0180, 0.0145, 0.0211, 0.0183, 0.0138], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 04:02:11,424 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 +2023-03-21 04:02:13,735 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 04:02:20,677 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-03-21 04:02:20,895 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=72623.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:02:24,307 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.326e+02 1.824e+02 2.165e+02 2.550e+02 4.522e+02, threshold=4.329e+02, percent-clipped=1.0 +2023-03-21 04:02:24,360 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 04:02:31,881 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 04:02:34,897 INFO [train.py:901] (1/2) Epoch 26, batch 2050, loss[loss=0.139, simple_loss=0.2167, pruned_loss=0.0306, over 7233.00 frames. ], tot_loss[loss=0.1416, simple_loss=0.2215, pruned_loss=0.03087, over 1442606.31 frames. ], batch size: 45, lr: 6.33e-03, grad_scale: 16.0 +2023-03-21 04:02:42,604 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-21 04:02:51,383 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72682.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:02:56,508 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8958, 2.7662, 2.9026, 2.9803, 2.4746, 2.5061, 2.9690, 2.1898], + device='cuda:1'), covar=tensor([0.0563, 0.0604, 0.0524, 0.0620, 0.0712, 0.0922, 0.0739, 0.1728], + device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0339, 0.0272, 0.0363, 0.0305, 0.0300, 0.0345, 0.0277], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:02:59,115 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4224, 1.0840, 1.6200, 1.7717, 1.5475, 1.8215, 1.4907, 1.7266], + device='cuda:1'), covar=tensor([0.3845, 0.3655, 0.2202, 0.1466, 0.1898, 0.1580, 0.2096, 0.2696], + device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0070, 0.0056, 0.0052, 0.0054, 0.0053, 0.0085, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:02:59,561 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1399, 4.6446, 4.7455, 4.6795, 4.6000, 4.1948, 4.7699, 4.5994], + device='cuda:1'), covar=tensor([0.0474, 0.0443, 0.0328, 0.0441, 0.0323, 0.0435, 0.0277, 0.0431], + device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0234, 0.0178, 0.0179, 0.0145, 0.0210, 0.0180, 0.0137], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 04:03:01,487 INFO [train.py:901] (1/2) Epoch 26, batch 2100, loss[loss=0.1594, simple_loss=0.2421, pruned_loss=0.03841, over 7134.00 frames. ], tot_loss[loss=0.141, simple_loss=0.2208, pruned_loss=0.03057, over 1441803.47 frames. ], batch size: 98, lr: 6.33e-03, grad_scale: 8.0 +2023-03-21 04:03:06,502 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 04:03:09,551 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 04:03:16,556 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.317e+02 1.881e+02 2.177e+02 2.542e+02 4.877e+02, threshold=4.354e+02, percent-clipped=2.0 +2023-03-21 04:03:20,557 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 +2023-03-21 04:03:27,372 INFO [train.py:901] (1/2) Epoch 26, batch 2150, loss[loss=0.1596, simple_loss=0.2379, pruned_loss=0.04068, over 7322.00 frames. ], tot_loss[loss=0.1414, simple_loss=0.2213, pruned_loss=0.03072, over 1443462.90 frames. ], batch size: 83, lr: 6.32e-03, grad_scale: 8.0 +2023-03-21 04:03:33,246 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.87 vs. limit=5.0 +2023-03-21 04:03:44,043 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-03-21 04:03:45,287 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72785.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:03:53,546 INFO [train.py:901] (1/2) Epoch 26, batch 2200, loss[loss=0.134, simple_loss=0.2146, pruned_loss=0.02669, over 7217.00 frames. ], tot_loss[loss=0.1409, simple_loss=0.2209, pruned_loss=0.03046, over 1444214.70 frames. ], batch size: 39, lr: 6.32e-03, grad_scale: 8.0 +2023-03-21 04:03:56,596 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 04:04:08,277 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9231, 3.8121, 3.1869, 3.3341, 2.7572, 2.1528, 1.7384, 3.8694], + device='cuda:1'), covar=tensor([0.0036, 0.0081, 0.0113, 0.0071, 0.0154, 0.0463, 0.0527, 0.0047], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0079, 0.0099, 0.0085, 0.0112, 0.0119, 0.0116, 0.0092], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 04:04:08,310 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5032, 1.1113, 1.4748, 1.7529, 1.4489, 1.7808, 1.3568, 1.7514], + device='cuda:1'), covar=tensor([0.2163, 0.4037, 0.1469, 0.0857, 0.1546, 0.1500, 0.1637, 0.1161], + device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0068, 0.0055, 0.0051, 0.0053, 0.0052, 0.0084, 0.0054], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:04:09,308 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.820e+02 2.213e+02 2.627e+02 5.046e+02, threshold=4.426e+02, percent-clipped=1.0 +2023-03-21 04:04:10,415 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=72833.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:04:19,360 INFO [train.py:901] (1/2) Epoch 26, batch 2250, loss[loss=0.1382, simple_loss=0.2171, pruned_loss=0.02969, over 7215.00 frames. ], tot_loss[loss=0.1417, simple_loss=0.2218, pruned_loss=0.03083, over 1443990.27 frames. ], batch size: 45, lr: 6.32e-03, grad_scale: 8.0 +2023-03-21 04:04:30,549 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 04:04:31,072 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 04:04:41,848 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72894.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:04:43,174 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 04:04:45,203 INFO [train.py:901] (1/2) Epoch 26, batch 2300, loss[loss=0.1288, simple_loss=0.221, pruned_loss=0.01829, over 7327.00 frames. ], tot_loss[loss=0.1413, simple_loss=0.2215, pruned_loss=0.03054, over 1444841.94 frames. ], batch size: 75, lr: 6.32e-03, grad_scale: 8.0 +2023-03-21 04:04:47,224 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-21 04:05:01,082 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.198e+02 1.738e+02 1.984e+02 2.359e+02 4.663e+02, threshold=3.968e+02, percent-clipped=1.0 +2023-03-21 04:05:11,646 INFO [train.py:901] (1/2) Epoch 26, batch 2350, loss[loss=0.1427, simple_loss=0.2319, pruned_loss=0.02679, over 7303.00 frames. ], tot_loss[loss=0.1413, simple_loss=0.2212, pruned_loss=0.03063, over 1442886.79 frames. ], batch size: 86, lr: 6.32e-03, grad_scale: 8.0 +2023-03-21 04:05:13,823 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72955.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:05:17,547 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-03-21 04:05:18,422 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72964.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 04:05:22,237 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0445, 4.1933, 3.9759, 4.2404, 3.8470, 4.1793, 4.4618, 4.5332], + device='cuda:1'), covar=tensor([0.0184, 0.0125, 0.0168, 0.0139, 0.0301, 0.0226, 0.0196, 0.0140], + device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0117, 0.0106, 0.0113, 0.0105, 0.0093, 0.0094, 0.0090], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:05:27,431 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72982.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:05:30,295 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 04:05:36,345 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 04:05:37,111 INFO [train.py:901] (1/2) Epoch 26, batch 2400, loss[loss=0.154, simple_loss=0.2334, pruned_loss=0.03732, over 7263.00 frames. ], tot_loss[loss=0.1417, simple_loss=0.2216, pruned_loss=0.03087, over 1443249.75 frames. ], batch size: 89, lr: 6.31e-03, grad_scale: 8.0 +2023-03-21 04:05:46,823 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 04:05:49,368 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 04:05:49,983 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73025.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 04:05:52,982 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=73030.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:05:53,416 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.197e+02 1.959e+02 2.204e+02 2.693e+02 9.099e+02, threshold=4.408e+02, percent-clipped=4.0 +2023-03-21 04:05:54,230 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-03-21 04:06:02,024 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.3158, 4.7344, 4.5710, 5.2244, 5.0382, 5.1270, 4.5263, 4.7961], + device='cuda:1'), covar=tensor([0.0690, 0.2436, 0.2259, 0.1036, 0.0830, 0.1207, 0.0796, 0.0956], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0365, 0.0281, 0.0282, 0.0207, 0.0349, 0.0211, 0.0257], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:06:03,420 INFO [train.py:901] (1/2) Epoch 26, batch 2450, loss[loss=0.1458, simple_loss=0.2325, pruned_loss=0.0296, over 7141.00 frames. ], tot_loss[loss=0.1417, simple_loss=0.2218, pruned_loss=0.03082, over 1443238.45 frames. ], batch size: 98, lr: 6.31e-03, grad_scale: 8.0 +2023-03-21 04:06:03,520 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9484, 4.0486, 3.8256, 4.0762, 3.6672, 4.0218, 4.3151, 4.3109], + device='cuda:1'), covar=tensor([0.0205, 0.0160, 0.0193, 0.0165, 0.0345, 0.0456, 0.0193, 0.0181], + device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0117, 0.0106, 0.0113, 0.0106, 0.0094, 0.0094, 0.0090], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:06:08,373 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 +2023-03-21 04:06:12,851 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-03-21 04:06:16,129 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 04:06:29,490 INFO [train.py:901] (1/2) Epoch 26, batch 2500, loss[loss=0.1454, simple_loss=0.2244, pruned_loss=0.03324, over 7316.00 frames. ], tot_loss[loss=0.1414, simple_loss=0.2214, pruned_loss=0.03067, over 1442648.01 frames. ], batch size: 80, lr: 6.31e-03, grad_scale: 8.0 +2023-03-21 04:06:39,364 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 +2023-03-21 04:06:42,135 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 04:06:45,137 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.825e+02 2.142e+02 2.392e+02 4.308e+02, threshold=4.285e+02, percent-clipped=0.0 +2023-03-21 04:06:51,663 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.64 vs. limit=5.0 +2023-03-21 04:06:55,366 INFO [train.py:901] (1/2) Epoch 26, batch 2550, loss[loss=0.1336, simple_loss=0.2198, pruned_loss=0.0237, over 7241.00 frames. ], tot_loss[loss=0.1418, simple_loss=0.222, pruned_loss=0.0308, over 1443742.82 frames. ], batch size: 89, lr: 6.31e-03, grad_scale: 8.0 +2023-03-21 04:07:21,555 INFO [train.py:901] (1/2) Epoch 26, batch 2600, loss[loss=0.1267, simple_loss=0.2122, pruned_loss=0.02062, over 7312.00 frames. ], tot_loss[loss=0.1412, simple_loss=0.2214, pruned_loss=0.03051, over 1444029.66 frames. ], batch size: 83, lr: 6.31e-03, grad_scale: 8.0 +2023-03-21 04:07:36,914 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.372e+02 1.805e+02 2.103e+02 2.646e+02 6.004e+02, threshold=4.206e+02, percent-clipped=1.0 +2023-03-21 04:07:45,318 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4785, 2.7778, 2.4172, 2.6852, 2.8119, 2.4648, 2.7119, 2.5206], + device='cuda:1'), covar=tensor([0.0637, 0.1002, 0.1524, 0.1238, 0.0633, 0.0608, 0.0783, 0.1065], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0052, 0.0060, 0.0053, 0.0050, 0.0054, 0.0052, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:07:46,243 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73250.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:07:46,657 INFO [train.py:901] (1/2) Epoch 26, batch 2650, loss[loss=0.1459, simple_loss=0.2321, pruned_loss=0.02981, over 7263.00 frames. ], tot_loss[loss=0.1417, simple_loss=0.2217, pruned_loss=0.03084, over 1441265.57 frames. ], batch size: 52, lr: 6.30e-03, grad_scale: 8.0 +2023-03-21 04:07:56,666 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3551, 2.9301, 2.6045, 2.7682, 2.9528, 2.3599, 2.7079, 2.5626], + device='cuda:1'), covar=tensor([0.1740, 0.1192, 0.1430, 0.1628, 0.1014, 0.1043, 0.1063, 0.1381], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0051, 0.0060, 0.0053, 0.0050, 0.0054, 0.0052, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:08:07,515 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73293.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:08:11,327 INFO [train.py:901] (1/2) Epoch 26, batch 2700, loss[loss=0.1422, simple_loss=0.2246, pruned_loss=0.02988, over 7273.00 frames. ], tot_loss[loss=0.1405, simple_loss=0.2203, pruned_loss=0.03034, over 1440930.50 frames. ], batch size: 89, lr: 6.30e-03, grad_scale: 8.0 +2023-03-21 04:08:20,751 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73320.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 04:08:26,064 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.284e+02 1.768e+02 2.025e+02 2.508e+02 4.489e+02, threshold=4.051e+02, percent-clipped=2.0 +2023-03-21 04:08:27,207 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7656, 2.3291, 2.1327, 1.9196, 2.0651, 1.8154, 1.7325, 1.6624], + device='cuda:1'), covar=tensor([0.0381, 0.0229, 0.0276, 0.0266, 0.0399, 0.0367, 0.0290, 0.0279], + device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0031, 0.0032, 0.0030, 0.0030, 0.0031, 0.0034, 0.0033], + device='cuda:1'), out_proj_covar=tensor([8.1981e-05, 8.0926e-05, 8.0134e-05, 7.7116e-05, 7.9334e-05, 7.9680e-05, + 8.4915e-05, 8.4729e-05], device='cuda:1') +2023-03-21 04:08:36,030 INFO [train.py:901] (1/2) Epoch 26, batch 2750, loss[loss=0.1447, simple_loss=0.2206, pruned_loss=0.03439, over 7273.00 frames. ], tot_loss[loss=0.141, simple_loss=0.221, pruned_loss=0.03052, over 1442683.25 frames. ], batch size: 57, lr: 6.30e-03, grad_scale: 8.0 +2023-03-21 04:08:37,651 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73354.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:08:41,674 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4660, 1.2372, 1.6125, 1.8426, 1.5825, 1.6959, 1.4569, 1.7675], + device='cuda:1'), covar=tensor([0.2687, 0.4656, 0.1456, 0.1516, 0.1182, 0.1803, 0.2056, 0.2684], + device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0070, 0.0057, 0.0052, 0.0054, 0.0055, 0.0086, 0.0055], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:08:44,154 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0010, 2.4260, 2.2428, 1.9979, 2.1457, 1.9955, 1.7601, 1.7165], + device='cuda:1'), covar=tensor([0.0282, 0.0160, 0.0320, 0.0245, 0.0272, 0.0489, 0.0310, 0.0287], + device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0031, 0.0032, 0.0030, 0.0030, 0.0031, 0.0034, 0.0033], + device='cuda:1'), out_proj_covar=tensor([8.2468e-05, 8.1150e-05, 8.0240e-05, 7.7279e-05, 7.9562e-05, 7.9757e-05, + 8.5086e-05, 8.4878e-05], device='cuda:1') +2023-03-21 04:08:51,491 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7850, 5.3505, 5.4398, 5.3274, 5.0587, 4.7944, 5.4501, 5.1570], + device='cuda:1'), covar=tensor([0.0482, 0.0375, 0.0318, 0.0446, 0.0364, 0.0357, 0.0282, 0.0481], + device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0239, 0.0181, 0.0182, 0.0147, 0.0213, 0.0185, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 04:09:00,996 INFO [train.py:901] (1/2) Epoch 26, batch 2800, loss[loss=0.1102, simple_loss=0.1839, pruned_loss=0.01829, over 6944.00 frames. ], tot_loss[loss=0.1406, simple_loss=0.2202, pruned_loss=0.0305, over 1441531.02 frames. ], batch size: 35, lr: 6.30e-03, grad_scale: 8.0 +2023-03-21 04:09:25,983 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 04:09:26,040 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 04:09:26,065 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 04:09:26,080 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 04:09:26,255 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 04:09:26,270 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 04:09:26,281 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 04:09:26,684 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 04:09:26,758 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 04:09:26,772 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 04:09:26,789 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 04:09:26,954 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 04:09:26,961 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 04:09:26,968 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 04:09:27,104 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 04:09:27,107 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 04:09:27,167 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 04:09:32,495 INFO [train.py:901] (1/2) Epoch 27, batch 0, loss[loss=0.1218, simple_loss=0.1842, pruned_loss=0.02975, over 6486.00 frames. ], tot_loss[loss=0.1218, simple_loss=0.1842, pruned_loss=0.02975, over 6486.00 frames. ], batch size: 28, lr: 6.18e-03, grad_scale: 8.0 +2023-03-21 04:09:32,496 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 04:09:39,395 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5543, 1.2087, 1.5268, 1.7932, 1.5347, 1.7777, 1.3506, 1.8745], + device='cuda:1'), covar=tensor([0.2005, 0.4117, 0.0955, 0.0952, 0.2084, 0.1108, 0.1844, 0.1320], + device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0069, 0.0056, 0.0051, 0.0053, 0.0054, 0.0084, 0.0054], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:09:42,510 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7813, 4.3372, 4.1188, 4.7684, 4.5623, 4.7998, 4.2795, 4.6144], + device='cuda:1'), covar=tensor([0.0585, 0.2420, 0.1900, 0.1269, 0.0895, 0.1018, 0.0611, 0.0685], + device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0359, 0.0277, 0.0277, 0.0205, 0.0345, 0.0208, 0.0255], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:09:57,908 INFO [train.py:935] (1/2) Epoch 27, validation: loss=0.1647, simple_loss=0.2543, pruned_loss=0.03752, over 1622729.00 frames. +2023-03-21 04:09:57,909 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 04:10:00,980 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.354e+02 1.845e+02 2.089e+02 2.471e+02 4.051e+02, threshold=4.178e+02, percent-clipped=1.0 +2023-03-21 04:10:05,665 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 04:10:16,119 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 04:10:23,303 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 04:10:24,286 INFO [train.py:901] (1/2) Epoch 27, batch 50, loss[loss=0.1456, simple_loss=0.2295, pruned_loss=0.03082, over 7284.00 frames. ], tot_loss[loss=0.1394, simple_loss=0.2195, pruned_loss=0.02969, over 325767.34 frames. ], batch size: 70, lr: 6.18e-03, grad_scale: 8.0 +2023-03-21 04:10:25,305 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 04:10:28,303 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 04:10:28,933 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5519, 1.2111, 1.7178, 1.8816, 1.6326, 1.8173, 1.5515, 1.8246], + device='cuda:1'), covar=tensor([0.2175, 0.3337, 0.0798, 0.0763, 0.2601, 0.1192, 0.1408, 0.2582], + device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0070, 0.0056, 0.0051, 0.0054, 0.0054, 0.0085, 0.0054], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:10:44,377 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 04:10:44,888 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 04:10:45,472 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6206, 3.6781, 3.5117, 3.5571, 2.8095, 3.3132, 3.5733, 3.3164], + device='cuda:1'), covar=tensor([0.0282, 0.0197, 0.0181, 0.0224, 0.0846, 0.0183, 0.0290, 0.0244], + device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0094, 0.0093, 0.0083, 0.0162, 0.0104, 0.0096, 0.0103], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:10:50,029 INFO [train.py:901] (1/2) Epoch 27, batch 100, loss[loss=0.1163, simple_loss=0.2015, pruned_loss=0.01551, over 7114.00 frames. ], tot_loss[loss=0.1392, simple_loss=0.2191, pruned_loss=0.02966, over 573541.27 frames. ], batch size: 41, lr: 6.18e-03, grad_scale: 8.0 +2023-03-21 04:10:52,944 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.895e+02 2.298e+02 2.662e+02 5.511e+02, threshold=4.595e+02, percent-clipped=3.0 +2023-03-21 04:11:00,995 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0170, 2.6320, 2.5764, 2.4267, 2.3034, 1.8908, 2.0532, 2.0941], + device='cuda:1'), covar=tensor([0.0582, 0.0280, 0.0311, 0.0214, 0.0475, 0.0696, 0.0445, 0.0239], + device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0031, 0.0032, 0.0030, 0.0030, 0.0031, 0.0034, 0.0033], + device='cuda:1'), out_proj_covar=tensor([8.1343e-05, 8.0639e-05, 7.9446e-05, 7.6623e-05, 7.8795e-05, 7.8669e-05, + 8.4369e-05, 8.4426e-05], device='cuda:1') +2023-03-21 04:11:02,484 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73550.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:11:15,579 INFO [train.py:901] (1/2) Epoch 27, batch 150, loss[loss=0.1404, simple_loss=0.2253, pruned_loss=0.02771, over 7313.00 frames. ], tot_loss[loss=0.1395, simple_loss=0.2201, pruned_loss=0.02947, over 766735.63 frames. ], batch size: 80, lr: 6.17e-03, grad_scale: 8.0 +2023-03-21 04:11:20,721 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5642, 3.5876, 2.7222, 3.1220, 2.4322, 2.0577, 1.6676, 3.6201], + device='cuda:1'), covar=tensor([0.0065, 0.0064, 0.0236, 0.0104, 0.0276, 0.0663, 0.0721, 0.0065], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0081, 0.0104, 0.0088, 0.0116, 0.0124, 0.0123, 0.0095], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 04:11:21,814 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5652, 3.4190, 2.3554, 3.9163, 2.8723, 3.2467, 1.7614, 2.4438], + device='cuda:1'), covar=tensor([0.0418, 0.0646, 0.2288, 0.0389, 0.0363, 0.0518, 0.3343, 0.1775], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0253, 0.0292, 0.0264, 0.0268, 0.0266, 0.0249, 0.0266], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:11:23,265 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73590.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:11:27,199 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=73598.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:11:39,085 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73620.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 04:11:41,453 INFO [train.py:901] (1/2) Epoch 27, batch 200, loss[loss=0.1391, simple_loss=0.2173, pruned_loss=0.03045, over 7316.00 frames. ], tot_loss[loss=0.1399, simple_loss=0.2197, pruned_loss=0.03, over 914393.94 frames. ], batch size: 59, lr: 6.17e-03, grad_scale: 8.0 +2023-03-21 04:11:44,516 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.178e+02 1.801e+02 2.024e+02 2.555e+02 5.461e+02, threshold=4.047e+02, percent-clipped=3.0 +2023-03-21 04:11:46,071 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 04:11:50,644 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 04:11:54,255 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73649.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:11:55,338 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73651.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:11:57,671 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 04:12:03,837 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=73668.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 04:12:03,932 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7921, 2.3326, 1.7113, 2.5061, 2.8240, 2.2002, 2.4208, 2.3352], + device='cuda:1'), covar=tensor([0.2025, 0.0860, 0.3460, 0.0729, 0.0246, 0.0188, 0.0299, 0.0360], + device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0234, 0.0259, 0.0262, 0.0180, 0.0181, 0.0207, 0.0216], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:12:07,148 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.52 vs. limit=5.0 +2023-03-21 04:12:07,319 INFO [train.py:901] (1/2) Epoch 27, batch 250, loss[loss=0.1285, simple_loss=0.2114, pruned_loss=0.02284, over 7292.00 frames. ], tot_loss[loss=0.1399, simple_loss=0.2195, pruned_loss=0.03011, over 1028828.46 frames. ], batch size: 77, lr: 6.17e-03, grad_scale: 8.0 +2023-03-21 04:12:10,285 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 04:12:19,309 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6459, 2.9413, 2.5603, 2.9465, 2.9096, 2.5483, 2.7949, 2.6249], + device='cuda:1'), covar=tensor([0.0742, 0.0717, 0.0848, 0.0813, 0.0824, 0.0633, 0.1362, 0.1085], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0051, 0.0059, 0.0052, 0.0050, 0.0053, 0.0051, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:12:31,140 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 04:12:32,612 INFO [train.py:901] (1/2) Epoch 27, batch 300, loss[loss=0.1337, simple_loss=0.2194, pruned_loss=0.02399, over 7363.00 frames. ], tot_loss[loss=0.14, simple_loss=0.2195, pruned_loss=0.03025, over 1120527.78 frames. ], batch size: 51, lr: 6.17e-03, grad_scale: 8.0 +2023-03-21 04:12:35,676 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.939e+02 2.170e+02 2.578e+02 5.056e+02, threshold=4.340e+02, percent-clipped=4.0 +2023-03-21 04:12:36,275 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73732.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:12:40,653 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 04:12:49,778 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.8466, 4.2310, 4.1765, 4.7541, 4.6434, 4.7124, 4.2849, 4.2999], + device='cuda:1'), covar=tensor([0.0669, 0.2362, 0.2105, 0.1032, 0.0823, 0.1164, 0.0729, 0.1212], + device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0355, 0.0273, 0.0274, 0.0203, 0.0340, 0.0206, 0.0252], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:12:52,317 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9719, 2.3689, 2.2562, 2.0581, 2.2380, 2.0983, 1.9982, 1.7344], + device='cuda:1'), covar=tensor([0.0465, 0.0342, 0.0307, 0.0320, 0.0514, 0.0341, 0.0300, 0.0325], + device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0031, 0.0031, 0.0029, 0.0030, 0.0031, 0.0033, 0.0033], + device='cuda:1'), out_proj_covar=tensor([8.0610e-05, 8.0206e-05, 7.8391e-05, 7.5476e-05, 7.7687e-05, 7.7765e-05, + 8.2834e-05, 8.3734e-05], device='cuda:1') +2023-03-21 04:12:58,077 INFO [train.py:901] (1/2) Epoch 27, batch 350, loss[loss=0.1359, simple_loss=0.2202, pruned_loss=0.02582, over 7274.00 frames. ], tot_loss[loss=0.1399, simple_loss=0.2197, pruned_loss=0.03003, over 1191999.74 frames. ], batch size: 52, lr: 6.17e-03, grad_scale: 8.0 +2023-03-21 04:13:07,867 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73793.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:13:13,501 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 04:13:24,629 INFO [train.py:901] (1/2) Epoch 27, batch 400, loss[loss=0.1459, simple_loss=0.2186, pruned_loss=0.03655, over 7265.00 frames. ], tot_loss[loss=0.1402, simple_loss=0.22, pruned_loss=0.0302, over 1245894.74 frames. ], batch size: 52, lr: 6.16e-03, grad_scale: 8.0 +2023-03-21 04:13:27,546 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.915e+02 2.214e+02 2.597e+02 4.451e+02, threshold=4.429e+02, percent-clipped=1.0 +2023-03-21 04:13:49,057 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-21 04:13:50,224 INFO [train.py:901] (1/2) Epoch 27, batch 450, loss[loss=0.1337, simple_loss=0.222, pruned_loss=0.02276, over 7272.00 frames. ], tot_loss[loss=0.14, simple_loss=0.2199, pruned_loss=0.03006, over 1291307.56 frames. ], batch size: 77, lr: 6.16e-03, grad_scale: 8.0 +2023-03-21 04:13:52,370 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1035, 2.7257, 2.1400, 2.9434, 3.2841, 2.9571, 2.9850, 2.5942], + device='cuda:1'), covar=tensor([0.2016, 0.0821, 0.3239, 0.0635, 0.0184, 0.0169, 0.0247, 0.0258], + device='cuda:1'), in_proj_covar=tensor([0.0260, 0.0232, 0.0258, 0.0262, 0.0179, 0.0180, 0.0207, 0.0215], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:13:54,673 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 04:13:55,140 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 04:14:15,704 INFO [train.py:901] (1/2) Epoch 27, batch 500, loss[loss=0.1093, simple_loss=0.1735, pruned_loss=0.02261, over 5793.00 frames. ], tot_loss[loss=0.1399, simple_loss=0.2193, pruned_loss=0.03018, over 1321925.43 frames. ], batch size: 25, lr: 6.16e-03, grad_scale: 8.0 +2023-03-21 04:14:18,674 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.251e+02 1.829e+02 2.169e+02 2.599e+02 4.226e+02, threshold=4.339e+02, percent-clipped=0.0 +2023-03-21 04:14:18,966 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 +2023-03-21 04:14:26,333 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73946.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:14:27,764 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 04:14:27,882 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73949.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:14:29,292 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 04:14:30,225 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 04:14:32,871 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 04:14:36,993 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 04:14:41,391 INFO [train.py:901] (1/2) Epoch 27, batch 550, loss[loss=0.1394, simple_loss=0.2276, pruned_loss=0.02555, over 7351.00 frames. ], tot_loss[loss=0.14, simple_loss=0.2196, pruned_loss=0.03019, over 1349471.02 frames. ], batch size: 63, lr: 6.16e-03, grad_scale: 8.0 +2023-03-21 04:14:44,722 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 +2023-03-21 04:14:49,396 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 04:14:53,077 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=73997.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:14:58,417 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 04:15:01,410 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 04:15:07,481 INFO [train.py:901] (1/2) Epoch 27, batch 600, loss[loss=0.1376, simple_loss=0.2151, pruned_loss=0.03, over 7231.00 frames. ], tot_loss[loss=0.1393, simple_loss=0.2192, pruned_loss=0.02975, over 1369531.47 frames. ], batch size: 55, lr: 6.15e-03, grad_scale: 8.0 +2023-03-21 04:15:08,020 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 04:15:10,483 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.222e+02 1.761e+02 2.044e+02 2.479e+02 4.557e+02, threshold=4.088e+02, percent-clipped=1.0 +2023-03-21 04:15:25,849 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 04:15:27,398 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74064.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:15:32,777 INFO [train.py:901] (1/2) Epoch 27, batch 650, loss[loss=0.1421, simple_loss=0.2275, pruned_loss=0.02838, over 7355.00 frames. ], tot_loss[loss=0.1395, simple_loss=0.2192, pruned_loss=0.02992, over 1385614.18 frames. ], batch size: 63, lr: 6.15e-03, grad_scale: 8.0 +2023-03-21 04:15:34,297 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 04:15:39,137 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 +2023-03-21 04:15:39,844 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74088.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:15:48,989 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 +2023-03-21 04:15:51,680 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 04:15:58,158 INFO [train.py:901] (1/2) Epoch 27, batch 700, loss[loss=0.1413, simple_loss=0.219, pruned_loss=0.03185, over 7335.00 frames. ], tot_loss[loss=0.1408, simple_loss=0.2202, pruned_loss=0.03064, over 1397071.82 frames. ], batch size: 51, lr: 6.15e-03, grad_scale: 8.0 +2023-03-21 04:15:58,323 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74125.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:15:59,681 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 04:16:01,737 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.376e+02 1.824e+02 2.144e+02 2.665e+02 4.628e+02, threshold=4.287e+02, percent-clipped=3.0 +2023-03-21 04:16:24,401 INFO [train.py:901] (1/2) Epoch 27, batch 750, loss[loss=0.1334, simple_loss=0.2187, pruned_loss=0.02401, over 7209.00 frames. ], tot_loss[loss=0.1407, simple_loss=0.22, pruned_loss=0.03066, over 1406021.17 frames. ], batch size: 93, lr: 6.15e-03, grad_scale: 8.0 +2023-03-21 04:16:24,952 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 04:16:25,429 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 04:16:27,022 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8311, 3.6520, 3.6575, 3.5548, 3.5125, 3.4849, 3.7791, 3.3949], + device='cuda:1'), covar=tensor([0.0174, 0.0166, 0.0122, 0.0209, 0.0454, 0.0126, 0.0169, 0.0180], + device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0094, 0.0091, 0.0082, 0.0160, 0.0101, 0.0095, 0.0101], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:16:29,711 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-03-21 04:16:40,834 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 04:16:45,345 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 04:16:50,414 INFO [train.py:901] (1/2) Epoch 27, batch 800, loss[loss=0.136, simple_loss=0.2221, pruned_loss=0.02495, over 7228.00 frames. ], tot_loss[loss=0.1404, simple_loss=0.2198, pruned_loss=0.03049, over 1415329.71 frames. ], batch size: 93, lr: 6.15e-03, grad_scale: 8.0 +2023-03-21 04:16:51,922 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 04:16:52,971 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 04:16:53,435 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.936e+02 2.182e+02 2.596e+02 4.514e+02, threshold=4.364e+02, percent-clipped=1.0 +2023-03-21 04:16:56,218 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-03-21 04:16:58,072 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74240.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:17:01,068 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74246.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:17:01,617 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7771, 3.0389, 2.5818, 3.8576, 1.8496, 3.6865, 1.5072, 3.1244], + device='cuda:1'), covar=tensor([0.0129, 0.0867, 0.1408, 0.0134, 0.3825, 0.0164, 0.1204, 0.0367], + device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0255, 0.0274, 0.0199, 0.0261, 0.0206, 0.0247, 0.0231], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:17:03,998 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 04:17:16,441 INFO [train.py:901] (1/2) Epoch 27, batch 850, loss[loss=0.1413, simple_loss=0.2216, pruned_loss=0.03046, over 7260.00 frames. ], tot_loss[loss=0.1406, simple_loss=0.2203, pruned_loss=0.03045, over 1423119.86 frames. ], batch size: 52, lr: 6.14e-03, grad_scale: 8.0 +2023-03-21 04:17:23,541 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 04:17:23,551 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 04:17:26,236 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=74294.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:17:28,737 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 04:17:29,888 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74301.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:17:32,893 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 04:17:41,146 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3501, 3.0733, 2.0823, 3.4804, 2.5342, 3.1059, 1.6092, 2.1782], + device='cuda:1'), covar=tensor([0.0356, 0.0811, 0.2460, 0.0525, 0.0529, 0.0618, 0.3198, 0.1662], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0253, 0.0289, 0.0265, 0.0269, 0.0265, 0.0247, 0.0268], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:17:42,435 INFO [train.py:901] (1/2) Epoch 27, batch 900, loss[loss=0.1321, simple_loss=0.2158, pruned_loss=0.02425, over 7329.00 frames. ], tot_loss[loss=0.141, simple_loss=0.2209, pruned_loss=0.03057, over 1428197.46 frames. ], batch size: 61, lr: 6.14e-03, grad_scale: 8.0 +2023-03-21 04:17:45,468 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.281e+02 1.810e+02 2.047e+02 2.519e+02 5.499e+02, threshold=4.094e+02, percent-clipped=2.0 +2023-03-21 04:18:08,103 INFO [train.py:901] (1/2) Epoch 27, batch 950, loss[loss=0.1393, simple_loss=0.2165, pruned_loss=0.03101, over 7272.00 frames. ], tot_loss[loss=0.1409, simple_loss=0.221, pruned_loss=0.03044, over 1432523.74 frames. ], batch size: 57, lr: 6.14e-03, grad_scale: 8.0 +2023-03-21 04:18:09,108 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 04:18:14,612 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74388.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:18:31,606 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74420.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:18:33,505 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 04:18:33,996 INFO [train.py:901] (1/2) Epoch 27, batch 1000, loss[loss=0.1382, simple_loss=0.2189, pruned_loss=0.02874, over 7327.00 frames. ], tot_loss[loss=0.1406, simple_loss=0.2208, pruned_loss=0.03027, over 1437443.03 frames. ], batch size: 59, lr: 6.14e-03, grad_scale: 8.0 +2023-03-21 04:18:36,120 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6956, 3.8197, 3.7708, 3.9625, 3.5936, 3.9032, 4.0843, 4.1794], + device='cuda:1'), covar=tensor([0.0245, 0.0192, 0.0211, 0.0204, 0.0408, 0.0405, 0.0281, 0.0172], + device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0120, 0.0110, 0.0117, 0.0108, 0.0095, 0.0097, 0.0093], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:18:37,002 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 1.886e+02 2.186e+02 2.588e+02 3.946e+02, threshold=4.372e+02, percent-clipped=0.0 +2023-03-21 04:18:40,168 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=74436.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:18:55,218 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 04:18:59,722 INFO [train.py:901] (1/2) Epoch 27, batch 1050, loss[loss=0.1411, simple_loss=0.2194, pruned_loss=0.03138, over 7365.00 frames. ], tot_loss[loss=0.1412, simple_loss=0.2211, pruned_loss=0.03061, over 1439609.05 frames. ], batch size: 73, lr: 6.14e-03, grad_scale: 8.0 +2023-03-21 04:19:16,205 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 04:19:20,854 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 04:19:24,140 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2507, 3.2404, 2.2962, 3.8146, 2.7495, 3.4119, 1.7588, 2.2832], + device='cuda:1'), covar=tensor([0.0406, 0.0505, 0.2940, 0.0561, 0.0452, 0.0501, 0.3816, 0.2189], + device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0251, 0.0292, 0.0267, 0.0270, 0.0265, 0.0247, 0.0268], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:19:25,133 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2920, 2.4429, 2.0736, 3.5426, 1.6797, 3.3164, 1.2685, 2.5914], + device='cuda:1'), covar=tensor([0.0166, 0.1267, 0.1790, 0.0151, 0.3957, 0.0232, 0.1397, 0.0321], + device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0257, 0.0275, 0.0199, 0.0262, 0.0208, 0.0247, 0.0232], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:19:25,978 INFO [train.py:901] (1/2) Epoch 27, batch 1100, loss[loss=0.1267, simple_loss=0.1997, pruned_loss=0.02685, over 7229.00 frames. ], tot_loss[loss=0.1413, simple_loss=0.2213, pruned_loss=0.03063, over 1441259.26 frames. ], batch size: 45, lr: 6.13e-03, grad_scale: 8.0 +2023-03-21 04:19:28,999 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.409e+02 1.926e+02 2.261e+02 2.649e+02 8.248e+02, threshold=4.522e+02, percent-clipped=2.0 +2023-03-21 04:19:30,983 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.12 vs. limit=5.0 +2023-03-21 04:19:50,604 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 04:19:51,082 INFO [train.py:901] (1/2) Epoch 27, batch 1150, loss[loss=0.1226, simple_loss=0.203, pruned_loss=0.02112, over 7216.00 frames. ], tot_loss[loss=0.1407, simple_loss=0.2206, pruned_loss=0.03035, over 1441820.10 frames. ], batch size: 50, lr: 6.13e-03, grad_scale: 8.0 +2023-03-21 04:19:51,097 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 04:20:01,846 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74595.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:20:02,290 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74596.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:20:03,730 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 04:20:04,222 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 04:20:08,113 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4244, 4.0482, 4.0718, 4.1437, 4.0370, 3.9659, 4.3211, 3.7599], + device='cuda:1'), covar=tensor([0.0125, 0.0154, 0.0126, 0.0133, 0.0411, 0.0129, 0.0144, 0.0224], + device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0093, 0.0092, 0.0081, 0.0160, 0.0100, 0.0095, 0.0102], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:20:10,919 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.35 vs. limit=5.0 +2023-03-21 04:20:17,607 INFO [train.py:901] (1/2) Epoch 27, batch 1200, loss[loss=0.1438, simple_loss=0.2273, pruned_loss=0.03018, over 7209.00 frames. ], tot_loss[loss=0.1404, simple_loss=0.2205, pruned_loss=0.03016, over 1443129.32 frames. ], batch size: 93, lr: 6.13e-03, grad_scale: 8.0 +2023-03-21 04:20:20,649 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.288e+02 1.763e+02 2.138e+02 2.575e+02 4.166e+02, threshold=4.276e+02, percent-clipped=0.0 +2023-03-21 04:20:31,025 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6401, 3.2150, 3.5238, 3.4532, 3.1719, 2.9757, 3.6973, 2.6455], + device='cuda:1'), covar=tensor([0.0336, 0.0456, 0.0428, 0.0527, 0.0596, 0.0748, 0.0446, 0.1495], + device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0345, 0.0273, 0.0365, 0.0302, 0.0302, 0.0346, 0.0276], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:20:33,512 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74656.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:20:37,459 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 04:20:42,916 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-21 04:20:43,647 INFO [train.py:901] (1/2) Epoch 27, batch 1250, loss[loss=0.1392, simple_loss=0.2253, pruned_loss=0.02656, over 7329.00 frames. ], tot_loss[loss=0.1401, simple_loss=0.2202, pruned_loss=0.03005, over 1443368.14 frames. ], batch size: 59, lr: 6.13e-03, grad_scale: 16.0 +2023-03-21 04:21:02,130 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 04:21:06,124 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 04:21:06,771 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74720.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:21:07,163 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 04:21:09,127 INFO [train.py:901] (1/2) Epoch 27, batch 1300, loss[loss=0.1552, simple_loss=0.2338, pruned_loss=0.03827, over 6766.00 frames. ], tot_loss[loss=0.1403, simple_loss=0.2197, pruned_loss=0.03039, over 1440236.95 frames. ], batch size: 107, lr: 6.13e-03, grad_scale: 16.0 +2023-03-21 04:21:12,184 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.400e+02 1.865e+02 2.147e+02 2.548e+02 5.558e+02, threshold=4.294e+02, percent-clipped=5.0 +2023-03-21 04:21:24,080 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3660, 1.2599, 1.6899, 1.9153, 1.5891, 1.9243, 1.5557, 1.7920], + device='cuda:1'), covar=tensor([0.2829, 0.3801, 0.1658, 0.1116, 0.1721, 0.1239, 0.1447, 0.2174], + device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0069, 0.0055, 0.0050, 0.0053, 0.0053, 0.0084, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:21:29,407 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 04:21:31,475 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=74768.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:21:31,964 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 04:21:34,996 INFO [train.py:901] (1/2) Epoch 27, batch 1350, loss[loss=0.176, simple_loss=0.249, pruned_loss=0.05154, over 6657.00 frames. ], tot_loss[loss=0.1403, simple_loss=0.2198, pruned_loss=0.03035, over 1441017.28 frames. ], batch size: 107, lr: 6.12e-03, grad_scale: 16.0 +2023-03-21 04:21:35,013 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 04:21:45,702 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 04:21:54,635 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2283, 2.9417, 2.1666, 3.6180, 2.5893, 3.0942, 1.5195, 2.2171], + device='cuda:1'), covar=tensor([0.0394, 0.1085, 0.2541, 0.0422, 0.0473, 0.0606, 0.3340, 0.1771], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0254, 0.0293, 0.0267, 0.0270, 0.0267, 0.0248, 0.0270], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:22:00,958 INFO [train.py:901] (1/2) Epoch 27, batch 1400, loss[loss=0.1583, simple_loss=0.2467, pruned_loss=0.03496, over 7152.00 frames. ], tot_loss[loss=0.1403, simple_loss=0.2202, pruned_loss=0.03017, over 1441444.25 frames. ], batch size: 98, lr: 6.12e-03, grad_scale: 16.0 +2023-03-21 04:22:03,882 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.860e+02 2.142e+02 2.591e+02 4.616e+02, threshold=4.284e+02, percent-clipped=2.0 +2023-03-21 04:22:06,726 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 +2023-03-21 04:22:17,639 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 04:22:27,219 INFO [train.py:901] (1/2) Epoch 27, batch 1450, loss[loss=0.1459, simple_loss=0.2214, pruned_loss=0.03525, over 7304.00 frames. ], tot_loss[loss=0.1401, simple_loss=0.22, pruned_loss=0.03014, over 1441371.86 frames. ], batch size: 59, lr: 6.12e-03, grad_scale: 16.0 +2023-03-21 04:22:37,815 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74896.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:22:41,741 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 04:22:44,408 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74909.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:22:51,576 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 +2023-03-21 04:22:52,252 INFO [train.py:901] (1/2) Epoch 27, batch 1500, loss[loss=0.1532, simple_loss=0.2325, pruned_loss=0.03697, over 7263.00 frames. ], tot_loss[loss=0.1405, simple_loss=0.2206, pruned_loss=0.03025, over 1442589.23 frames. ], batch size: 70, lr: 6.12e-03, grad_scale: 16.0 +2023-03-21 04:22:55,753 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.463e+02 1.784e+02 2.140e+02 2.549e+02 4.343e+02, threshold=4.279e+02, percent-clipped=1.0 +2023-03-21 04:22:57,325 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 04:23:02,467 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=74944.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:23:04,608 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3654, 1.5394, 1.4520, 1.4358, 1.5852, 1.4534, 1.4370, 1.1374], + device='cuda:1'), covar=tensor([0.0175, 0.0157, 0.0171, 0.0162, 0.0103, 0.0093, 0.0126, 0.0129], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0028, 0.0028, 0.0030, 0.0028, 0.0027, 0.0030, 0.0038], + device='cuda:1'), out_proj_covar=tensor([3.5142e-05, 3.1816e-05, 3.2532e-05, 3.3602e-05, 3.1481e-05, 3.0388e-05, + 3.3787e-05, 4.2436e-05], device='cuda:1') +2023-03-21 04:23:06,644 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74951.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:23:13,219 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74964.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:23:16,242 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74970.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:23:18,514 INFO [train.py:901] (1/2) Epoch 27, batch 1550, loss[loss=0.1644, simple_loss=0.242, pruned_loss=0.04343, over 7289.00 frames. ], tot_loss[loss=0.1404, simple_loss=0.2202, pruned_loss=0.03031, over 1439796.18 frames. ], batch size: 57, lr: 6.12e-03, grad_scale: 8.0 +2023-03-21 04:23:21,963 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 04:23:44,537 INFO [train.py:901] (1/2) Epoch 27, batch 1600, loss[loss=0.1436, simple_loss=0.2191, pruned_loss=0.03405, over 7330.00 frames. ], tot_loss[loss=0.1409, simple_loss=0.2206, pruned_loss=0.03062, over 1439102.50 frames. ], batch size: 83, lr: 6.11e-03, grad_scale: 8.0 +2023-03-21 04:23:44,699 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75025.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:23:48,018 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 1.920e+02 2.273e+02 2.923e+02 5.005e+02, threshold=4.545e+02, percent-clipped=3.0 +2023-03-21 04:23:48,158 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9985, 4.0530, 3.2539, 3.5802, 3.0030, 2.1904, 1.8223, 4.1552], + device='cuda:1'), covar=tensor([0.0057, 0.0072, 0.0143, 0.0083, 0.0166, 0.0528, 0.0607, 0.0047], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0082, 0.0104, 0.0089, 0.0118, 0.0125, 0.0124, 0.0097], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 04:23:49,166 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2037, 1.5834, 1.3501, 1.2580, 1.4280, 1.4073, 1.4543, 1.1915], + device='cuda:1'), covar=tensor([0.0142, 0.0090, 0.0137, 0.0131, 0.0084, 0.0063, 0.0099, 0.0109], + device='cuda:1'), in_proj_covar=tensor([0.0030, 0.0028, 0.0028, 0.0029, 0.0027, 0.0027, 0.0029, 0.0037], + device='cuda:1'), out_proj_covar=tensor([3.4674e-05, 3.1303e-05, 3.2247e-05, 3.2965e-05, 3.1029e-05, 2.9922e-05, + 3.2994e-05, 4.1760e-05], device='cuda:1') +2023-03-21 04:23:50,597 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4495, 1.3556, 1.8578, 2.0136, 1.6735, 1.9818, 1.7819, 1.9125], + device='cuda:1'), covar=tensor([0.2570, 0.3587, 0.2048, 0.1672, 0.1661, 0.2077, 0.1416, 0.3432], + device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0070, 0.0056, 0.0052, 0.0054, 0.0054, 0.0085, 0.0054], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:23:53,572 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 04:23:54,087 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 04:23:57,071 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 04:24:06,197 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 04:24:10,248 INFO [train.py:901] (1/2) Epoch 27, batch 1650, loss[loss=0.1528, simple_loss=0.2353, pruned_loss=0.03511, over 7217.00 frames. ], tot_loss[loss=0.1415, simple_loss=0.2214, pruned_loss=0.03082, over 1439190.82 frames. ], batch size: 50, lr: 6.11e-03, grad_scale: 8.0 +2023-03-21 04:24:11,254 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 04:24:15,941 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0838, 4.5451, 4.6444, 4.5999, 4.6286, 4.1643, 4.6892, 4.5790], + device='cuda:1'), covar=tensor([0.0519, 0.0466, 0.0496, 0.0492, 0.0314, 0.0440, 0.0403, 0.0464], + device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0242, 0.0185, 0.0184, 0.0146, 0.0216, 0.0190, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 04:24:19,335 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 04:24:20,609 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 +2023-03-21 04:24:36,048 INFO [train.py:901] (1/2) Epoch 27, batch 1700, loss[loss=0.1423, simple_loss=0.2299, pruned_loss=0.02734, over 7344.00 frames. ], tot_loss[loss=0.1405, simple_loss=0.2202, pruned_loss=0.03041, over 1438217.77 frames. ], batch size: 63, lr: 6.11e-03, grad_scale: 8.0 +2023-03-21 04:24:36,651 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.8969, 5.3501, 5.4123, 5.3674, 5.1977, 4.8975, 5.4843, 5.2634], + device='cuda:1'), covar=tensor([0.0381, 0.0344, 0.0409, 0.0475, 0.0312, 0.0358, 0.0282, 0.0422], + device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0241, 0.0184, 0.0184, 0.0146, 0.0217, 0.0190, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 04:24:37,540 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 04:24:40,086 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.205e+02 1.858e+02 2.103e+02 2.498e+02 6.232e+02, threshold=4.206e+02, percent-clipped=1.0 +2023-03-21 04:24:42,243 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 04:24:52,209 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 04:24:54,458 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 +2023-03-21 04:25:01,678 INFO [train.py:901] (1/2) Epoch 27, batch 1750, loss[loss=0.1529, simple_loss=0.2416, pruned_loss=0.03214, over 6679.00 frames. ], tot_loss[loss=0.141, simple_loss=0.2206, pruned_loss=0.03071, over 1436295.16 frames. ], batch size: 106, lr: 6.11e-03, grad_scale: 8.0 +2023-03-21 04:25:13,698 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 +2023-03-21 04:25:16,233 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 04:25:17,252 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 04:25:18,055 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-03-21 04:25:27,335 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8693, 2.2742, 2.8577, 2.8569, 2.8908, 2.7735, 2.3616, 2.8699], + device='cuda:1'), covar=tensor([0.1585, 0.0805, 0.1195, 0.1245, 0.0838, 0.1025, 0.2393, 0.1301], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0062, 0.0047, 0.0047, 0.0046, 0.0044, 0.0064, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:25:28,193 INFO [train.py:901] (1/2) Epoch 27, batch 1800, loss[loss=0.124, simple_loss=0.2041, pruned_loss=0.0219, over 7317.00 frames. ], tot_loss[loss=0.1403, simple_loss=0.2197, pruned_loss=0.03043, over 1437796.23 frames. ], batch size: 42, lr: 6.11e-03, grad_scale: 8.0 +2023-03-21 04:25:29,840 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75228.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:25:31,714 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.312e+02 1.830e+02 2.153e+02 2.628e+02 4.402e+02, threshold=4.306e+02, percent-clipped=1.0 +2023-03-21 04:25:39,341 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 04:25:41,392 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75251.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:25:48,527 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75265.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:25:48,559 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8453, 3.6115, 3.5960, 3.5177, 3.4754, 3.3497, 3.7064, 3.2920], + device='cuda:1'), covar=tensor([0.0122, 0.0155, 0.0126, 0.0185, 0.0441, 0.0140, 0.0166, 0.0198], + device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0094, 0.0092, 0.0081, 0.0160, 0.0100, 0.0095, 0.0102], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:25:51,063 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75270.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:25:52,725 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0014, 3.3642, 2.8348, 3.1536, 3.2428, 2.7226, 3.2441, 2.9812], + device='cuda:1'), covar=tensor([0.1018, 0.0964, 0.0977, 0.1622, 0.1515, 0.1098, 0.1193, 0.1715], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0053, 0.0060, 0.0054, 0.0052, 0.0055, 0.0052, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:25:53,626 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 04:25:54,101 INFO [train.py:901] (1/2) Epoch 27, batch 1850, loss[loss=0.1538, simple_loss=0.2383, pruned_loss=0.03468, over 7328.00 frames. ], tot_loss[loss=0.1394, simple_loss=0.2191, pruned_loss=0.0299, over 1439362.77 frames. ], batch size: 75, lr: 6.10e-03, grad_scale: 8.0 +2023-03-21 04:25:59,339 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0803, 2.4124, 2.4639, 2.1919, 2.2391, 2.0938, 1.8070, 1.8910], + device='cuda:1'), covar=tensor([0.0334, 0.0383, 0.0230, 0.0195, 0.0560, 0.0689, 0.0591, 0.0351], + device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0032, 0.0033, 0.0030, 0.0030, 0.0031, 0.0035, 0.0034], + device='cuda:1'), out_proj_covar=tensor([8.4020e-05, 8.2850e-05, 8.1849e-05, 7.7825e-05, 7.9477e-05, 8.0259e-05, + 8.6259e-05, 8.6868e-05], device='cuda:1') +2023-03-21 04:26:01,412 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75289.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:26:03,270 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 04:26:06,304 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=75299.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:26:17,400 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75320.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:26:17,525 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5953, 3.3392, 3.6857, 3.6352, 3.2774, 3.0363, 3.8038, 2.9049], + device='cuda:1'), covar=tensor([0.0299, 0.0356, 0.0391, 0.0378, 0.0573, 0.0787, 0.0434, 0.1384], + device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0345, 0.0272, 0.0361, 0.0299, 0.0300, 0.0343, 0.0275], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:26:18,980 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3525, 1.3160, 1.7675, 1.8811, 1.6015, 1.9704, 1.6217, 1.8780], + device='cuda:1'), covar=tensor([0.3515, 0.4357, 0.1681, 0.1537, 0.2467, 0.2625, 0.2019, 0.1539], + device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0072, 0.0057, 0.0053, 0.0054, 0.0055, 0.0087, 0.0055], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:26:19,786 INFO [train.py:901] (1/2) Epoch 27, batch 1900, loss[loss=0.1514, simple_loss=0.2213, pruned_loss=0.04071, over 7256.00 frames. ], tot_loss[loss=0.14, simple_loss=0.2194, pruned_loss=0.03025, over 1439002.55 frames. ], batch size: 64, lr: 6.10e-03, grad_scale: 8.0 +2023-03-21 04:26:20,325 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 04:26:23,012 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75331.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 04:26:23,336 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 1.826e+02 2.085e+02 2.530e+02 5.155e+02, threshold=4.170e+02, percent-clipped=3.0 +2023-03-21 04:26:43,964 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 04:26:45,474 INFO [train.py:901] (1/2) Epoch 27, batch 1950, loss[loss=0.1123, simple_loss=0.1763, pruned_loss=0.02417, over 6042.00 frames. ], tot_loss[loss=0.1402, simple_loss=0.2197, pruned_loss=0.03035, over 1438259.97 frames. ], batch size: 25, lr: 6.10e-03, grad_scale: 8.0 +2023-03-21 04:26:48,093 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0301, 3.7729, 3.7700, 3.7482, 3.6796, 3.5512, 3.8364, 3.4824], + device='cuda:1'), covar=tensor([0.0152, 0.0166, 0.0127, 0.0173, 0.0428, 0.0120, 0.0180, 0.0186], + device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0093, 0.0091, 0.0080, 0.0159, 0.0100, 0.0095, 0.0101], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:26:51,157 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3181, 3.4216, 3.3174, 3.3203, 3.2797, 3.4184, 3.5918, 3.6616], + device='cuda:1'), covar=tensor([0.0264, 0.0214, 0.0268, 0.0242, 0.0366, 0.0374, 0.0318, 0.0222], + device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0119, 0.0109, 0.0117, 0.0106, 0.0095, 0.0096, 0.0092], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:26:55,177 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 04:27:00,012 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 04:27:01,027 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 04:27:09,181 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3921, 1.2721, 1.6561, 1.8760, 1.5386, 1.9295, 1.5597, 1.7532], + device='cuda:1'), covar=tensor([0.1953, 0.3727, 0.1739, 0.1565, 0.1146, 0.1653, 0.1262, 0.2141], + device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0073, 0.0058, 0.0053, 0.0054, 0.0056, 0.0089, 0.0055], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:27:11,516 INFO [train.py:901] (1/2) Epoch 27, batch 2000, loss[loss=0.1285, simple_loss=0.2155, pruned_loss=0.02073, over 7257.00 frames. ], tot_loss[loss=0.1407, simple_loss=0.2204, pruned_loss=0.03049, over 1439074.43 frames. ], batch size: 55, lr: 6.10e-03, grad_scale: 8.0 +2023-03-21 04:27:14,942 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 1.802e+02 2.059e+02 2.426e+02 4.961e+02, threshold=4.118e+02, percent-clipped=2.0 +2023-03-21 04:27:16,512 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 04:27:19,085 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0926, 3.4776, 2.7998, 3.4578, 3.3494, 2.7416, 3.1982, 3.0738], + device='cuda:1'), covar=tensor([0.0495, 0.0580, 0.1087, 0.0661, 0.1311, 0.0837, 0.1032, 0.1496], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0052, 0.0060, 0.0053, 0.0051, 0.0055, 0.0051, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:27:24,241 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8086, 3.6379, 3.6652, 3.7611, 3.2940, 3.1656, 3.9471, 2.9401], + device='cuda:1'), covar=tensor([0.0430, 0.0403, 0.0366, 0.0457, 0.0606, 0.0711, 0.0420, 0.1375], + device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0342, 0.0270, 0.0360, 0.0298, 0.0298, 0.0343, 0.0273], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:27:28,184 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 04:27:32,463 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.41 vs. limit=5.0 +2023-03-21 04:27:34,067 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-03-21 04:27:35,250 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 04:27:37,895 INFO [train.py:901] (1/2) Epoch 27, batch 2050, loss[loss=0.1376, simple_loss=0.2204, pruned_loss=0.02736, over 7183.00 frames. ], tot_loss[loss=0.1403, simple_loss=0.22, pruned_loss=0.03027, over 1441405.37 frames. ], batch size: 98, lr: 6.10e-03, grad_scale: 8.0 +2023-03-21 04:27:40,520 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6635, 5.0960, 5.1503, 5.1250, 4.9336, 4.6730, 5.1624, 4.9389], + device='cuda:1'), covar=tensor([0.0455, 0.0368, 0.0364, 0.0419, 0.0309, 0.0340, 0.0333, 0.0438], + device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0241, 0.0184, 0.0185, 0.0146, 0.0217, 0.0191, 0.0140], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 04:28:03,074 INFO [train.py:901] (1/2) Epoch 27, batch 2100, loss[loss=0.1109, simple_loss=0.1948, pruned_loss=0.01352, over 7209.00 frames. ], tot_loss[loss=0.1394, simple_loss=0.2189, pruned_loss=0.0299, over 1441238.32 frames. ], batch size: 39, lr: 6.09e-03, grad_scale: 8.0 +2023-03-21 04:28:07,140 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.255e+02 1.775e+02 2.143e+02 2.491e+02 4.702e+02, threshold=4.286e+02, percent-clipped=1.0 +2023-03-21 04:28:08,678 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 04:28:11,592 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 04:28:24,249 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75565.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:28:29,118 INFO [train.py:901] (1/2) Epoch 27, batch 2150, loss[loss=0.1412, simple_loss=0.2198, pruned_loss=0.03125, over 7276.00 frames. ], tot_loss[loss=0.1395, simple_loss=0.2193, pruned_loss=0.02987, over 1442967.88 frames. ], batch size: 57, lr: 6.09e-03, grad_scale: 8.0 +2023-03-21 04:28:33,725 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75584.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:28:48,773 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3813, 4.1830, 4.3629, 4.3198, 4.4946, 4.5560, 4.3467, 4.3498], + device='cuda:1'), covar=tensor([0.0033, 0.0073, 0.0034, 0.0042, 0.0031, 0.0033, 0.0034, 0.0048], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0064, 0.0054, 0.0052, 0.0051, 0.0056, 0.0048, 0.0070], + device='cuda:1'), out_proj_covar=tensor([8.3097e-05, 1.4117e-04, 1.0932e-04, 9.9672e-05, 9.5194e-05, 1.0559e-04, + 1.0006e-04, 1.3886e-04], device='cuda:1') +2023-03-21 04:28:49,229 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=75613.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:28:52,777 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75620.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:28:55,172 INFO [train.py:901] (1/2) Epoch 27, batch 2200, loss[loss=0.1505, simple_loss=0.2309, pruned_loss=0.03507, over 7269.00 frames. ], tot_loss[loss=0.1396, simple_loss=0.2194, pruned_loss=0.02986, over 1441332.49 frames. ], batch size: 52, lr: 6.09e-03, grad_scale: 8.0 +2023-03-21 04:28:55,744 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75626.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 04:28:58,677 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.743e+02 2.140e+02 2.621e+02 4.761e+02, threshold=4.279e+02, percent-clipped=4.0 +2023-03-21 04:28:58,716 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 04:29:15,841 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6194, 3.4860, 3.4482, 3.3583, 3.3332, 3.2090, 3.5429, 3.2561], + device='cuda:1'), covar=tensor([0.0160, 0.0210, 0.0142, 0.0217, 0.0419, 0.0145, 0.0175, 0.0183], + device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0093, 0.0091, 0.0080, 0.0158, 0.0100, 0.0095, 0.0101], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:29:17,308 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=75668.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:29:20,768 INFO [train.py:901] (1/2) Epoch 27, batch 2250, loss[loss=0.1291, simple_loss=0.2093, pruned_loss=0.02448, over 7326.00 frames. ], tot_loss[loss=0.1404, simple_loss=0.2204, pruned_loss=0.03022, over 1439468.07 frames. ], batch size: 61, lr: 6.09e-03, grad_scale: 8.0 +2023-03-21 04:29:31,885 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 04:29:31,897 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 04:29:44,957 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 04:29:46,395 INFO [train.py:901] (1/2) Epoch 27, batch 2300, loss[loss=0.1419, simple_loss=0.215, pruned_loss=0.03435, over 7268.00 frames. ], tot_loss[loss=0.1401, simple_loss=0.22, pruned_loss=0.03007, over 1440562.91 frames. ], batch size: 77, lr: 6.09e-03, grad_scale: 8.0 +2023-03-21 04:29:48,352 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-21 04:29:50,597 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.180e+02 1.982e+02 2.222e+02 2.648e+02 6.200e+02, threshold=4.445e+02, percent-clipped=3.0 +2023-03-21 04:30:03,511 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8468, 2.6289, 2.3039, 3.8826, 1.8316, 3.6764, 1.4618, 3.0241], + device='cuda:1'), covar=tensor([0.0151, 0.1154, 0.1767, 0.0183, 0.3845, 0.0216, 0.1259, 0.0396], + device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0254, 0.0276, 0.0198, 0.0261, 0.0209, 0.0244, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:30:12,344 INFO [train.py:901] (1/2) Epoch 27, batch 2350, loss[loss=0.1465, simple_loss=0.2287, pruned_loss=0.03218, over 7269.00 frames. ], tot_loss[loss=0.1405, simple_loss=0.2205, pruned_loss=0.03023, over 1441498.78 frames. ], batch size: 47, lr: 6.08e-03, grad_scale: 8.0 +2023-03-21 04:30:17,049 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.40 vs. limit=5.0 +2023-03-21 04:30:22,183 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 +2023-03-21 04:30:32,283 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 04:30:38,953 INFO [train.py:901] (1/2) Epoch 27, batch 2400, loss[loss=0.1415, simple_loss=0.2244, pruned_loss=0.02929, over 7340.00 frames. ], tot_loss[loss=0.141, simple_loss=0.221, pruned_loss=0.03051, over 1443919.97 frames. ], batch size: 54, lr: 6.08e-03, grad_scale: 8.0 +2023-03-21 04:30:39,975 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 04:30:42,374 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.255e+02 1.828e+02 2.153e+02 2.703e+02 5.254e+02, threshold=4.305e+02, percent-clipped=1.0 +2023-03-21 04:30:51,121 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 04:30:53,305 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2860, 1.5361, 1.3598, 1.4171, 1.5302, 1.4080, 1.3678, 1.1393], + device='cuda:1'), covar=tensor([0.0133, 0.0101, 0.0166, 0.0127, 0.0113, 0.0121, 0.0109, 0.0149], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0028, 0.0028, 0.0029, 0.0028, 0.0027, 0.0030, 0.0038], + device='cuda:1'), out_proj_covar=tensor([3.5279e-05, 3.1618e-05, 3.2289e-05, 3.2864e-05, 3.1566e-05, 3.0795e-05, + 3.3865e-05, 4.2344e-05], device='cuda:1') +2023-03-21 04:30:53,651 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 04:30:58,486 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 +2023-03-21 04:31:04,294 INFO [train.py:901] (1/2) Epoch 27, batch 2450, loss[loss=0.1153, simple_loss=0.1981, pruned_loss=0.01621, over 7118.00 frames. ], tot_loss[loss=0.1408, simple_loss=0.2207, pruned_loss=0.03042, over 1441753.04 frames. ], batch size: 41, lr: 6.08e-03, grad_scale: 8.0 +2023-03-21 04:31:09,799 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75884.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:31:21,388 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 04:31:24,103 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7932, 2.3611, 3.0526, 2.8942, 2.8285, 2.7258, 2.4146, 2.9350], + device='cuda:1'), covar=tensor([0.1456, 0.0656, 0.0881, 0.0865, 0.0942, 0.0885, 0.1838, 0.0974], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0062, 0.0047, 0.0046, 0.0045, 0.0044, 0.0063, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:31:30,993 INFO [train.py:901] (1/2) Epoch 27, batch 2500, loss[loss=0.1385, simple_loss=0.2198, pruned_loss=0.02863, over 7356.00 frames. ], tot_loss[loss=0.1401, simple_loss=0.22, pruned_loss=0.03014, over 1442764.97 frames. ], batch size: 63, lr: 6.08e-03, grad_scale: 8.0 +2023-03-21 04:31:31,608 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75926.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 04:31:32,637 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6614, 1.6957, 2.0231, 2.3931, 1.8460, 2.4153, 2.1626, 2.3150], + device='cuda:1'), covar=tensor([0.4931, 0.4125, 0.1458, 0.1301, 0.2511, 0.3588, 0.1644, 0.2046], + device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0072, 0.0058, 0.0053, 0.0055, 0.0056, 0.0089, 0.0055], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:31:34,496 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.362e+02 1.829e+02 2.210e+02 2.665e+02 4.876e+02, threshold=4.419e+02, percent-clipped=2.0 +2023-03-21 04:31:34,576 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=75932.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:31:46,183 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 04:31:48,776 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7905, 2.1062, 2.2295, 1.9505, 2.1789, 2.0272, 1.6213, 1.5486], + device='cuda:1'), covar=tensor([0.0349, 0.0333, 0.0175, 0.0168, 0.0218, 0.0337, 0.0416, 0.0370], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0032, 0.0033, 0.0030, 0.0031, 0.0031, 0.0035, 0.0035], + device='cuda:1'), out_proj_covar=tensor([8.5165e-05, 8.3169e-05, 8.2299e-05, 7.8000e-05, 8.1967e-05, 8.0148e-05, + 8.6814e-05, 8.7794e-05], device='cuda:1') +2023-03-21 04:31:56,287 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=75974.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:31:56,721 INFO [train.py:901] (1/2) Epoch 27, batch 2550, loss[loss=0.1322, simple_loss=0.2077, pruned_loss=0.02831, over 7268.00 frames. ], tot_loss[loss=0.1408, simple_loss=0.2205, pruned_loss=0.03052, over 1440911.32 frames. ], batch size: 47, lr: 6.08e-03, grad_scale: 8.0 +2023-03-21 04:31:57,878 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0478, 3.1515, 2.4888, 4.1255, 2.0114, 3.8029, 1.5760, 3.2694], + device='cuda:1'), covar=tensor([0.0107, 0.0868, 0.1490, 0.0160, 0.3521, 0.0209, 0.1178, 0.0382], + device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0255, 0.0278, 0.0199, 0.0261, 0.0210, 0.0245, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:32:03,935 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0252, 2.2085, 2.3006, 1.9414, 2.2330, 2.0121, 1.7595, 1.7849], + device='cuda:1'), covar=tensor([0.0194, 0.0457, 0.0170, 0.0216, 0.0362, 0.0460, 0.0256, 0.0234], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0032, 0.0033, 0.0030, 0.0031, 0.0031, 0.0035, 0.0035], + device='cuda:1'), out_proj_covar=tensor([8.4573e-05, 8.2508e-05, 8.1485e-05, 7.7423e-05, 8.1262e-05, 7.9432e-05, + 8.5678e-05, 8.7021e-05], device='cuda:1') +2023-03-21 04:32:26,510 INFO [train.py:901] (1/2) Epoch 27, batch 2600, loss[loss=0.1243, simple_loss=0.2018, pruned_loss=0.02343, over 7334.00 frames. ], tot_loss[loss=0.1408, simple_loss=0.2204, pruned_loss=0.03059, over 1440978.87 frames. ], batch size: 42, lr: 6.07e-03, grad_scale: 8.0 +2023-03-21 04:32:29,790 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.79 vs. limit=5.0 +2023-03-21 04:32:29,977 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.271e+02 1.838e+02 2.095e+02 2.590e+02 5.003e+02, threshold=4.189e+02, percent-clipped=2.0 +2023-03-21 04:32:51,385 INFO [train.py:901] (1/2) Epoch 27, batch 2650, loss[loss=0.1384, simple_loss=0.2153, pruned_loss=0.03077, over 7308.00 frames. ], tot_loss[loss=0.1404, simple_loss=0.2198, pruned_loss=0.03043, over 1440881.57 frames. ], batch size: 86, lr: 6.07e-03, grad_scale: 8.0 +2023-03-21 04:33:03,901 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2933, 1.7574, 1.5268, 1.5613, 1.7031, 1.4832, 1.5930, 1.0593], + device='cuda:1'), covar=tensor([0.0231, 0.0206, 0.0145, 0.0146, 0.0129, 0.0127, 0.0186, 0.0178], + device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0028, 0.0028, 0.0030, 0.0028, 0.0028, 0.0030, 0.0038], + device='cuda:1'), out_proj_covar=tensor([3.5810e-05, 3.2273e-05, 3.2347e-05, 3.3156e-05, 3.2079e-05, 3.1328e-05, + 3.4225e-05, 4.2678e-05], device='cuda:1') +2023-03-21 04:33:16,555 INFO [train.py:901] (1/2) Epoch 27, batch 2700, loss[loss=0.1405, simple_loss=0.2202, pruned_loss=0.0304, over 7318.00 frames. ], tot_loss[loss=0.1407, simple_loss=0.2205, pruned_loss=0.03046, over 1442661.51 frames. ], batch size: 83, lr: 6.07e-03, grad_scale: 8.0 +2023-03-21 04:33:19,921 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 1.823e+02 2.064e+02 2.639e+02 4.176e+02, threshold=4.127e+02, percent-clipped=0.0 +2023-03-21 04:33:35,571 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 +2023-03-21 04:33:41,160 INFO [train.py:901] (1/2) Epoch 27, batch 2750, loss[loss=0.1404, simple_loss=0.2237, pruned_loss=0.02855, over 7137.00 frames. ], tot_loss[loss=0.141, simple_loss=0.2207, pruned_loss=0.03065, over 1442989.24 frames. ], batch size: 98, lr: 6.07e-03, grad_scale: 8.0 +2023-03-21 04:34:00,195 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.69 vs. limit=5.0 +2023-03-21 04:34:02,671 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-21 04:34:06,260 INFO [train.py:901] (1/2) Epoch 27, batch 2800, loss[loss=0.1518, simple_loss=0.2329, pruned_loss=0.03542, over 7274.00 frames. ], tot_loss[loss=0.1408, simple_loss=0.2204, pruned_loss=0.03053, over 1442330.35 frames. ], batch size: 57, lr: 6.07e-03, grad_scale: 8.0 +2023-03-21 04:34:09,662 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 1.810e+02 2.119e+02 2.567e+02 4.960e+02, threshold=4.237e+02, percent-clipped=2.0 +2023-03-21 04:34:31,073 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 04:34:31,127 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 04:34:31,152 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 04:34:31,168 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 04:34:31,337 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 04:34:31,353 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 04:34:31,363 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 04:34:31,460 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 04:34:31,534 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 04:34:31,549 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 04:34:31,565 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 04:34:32,071 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 04:34:32,076 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 04:34:32,083 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 04:34:32,224 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 04:34:32,227 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 04:34:32,289 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 04:34:40,232 INFO [train.py:901] (1/2) Epoch 28, batch 0, loss[loss=0.1283, simple_loss=0.213, pruned_loss=0.02178, over 7346.00 frames. ], tot_loss[loss=0.1283, simple_loss=0.213, pruned_loss=0.02178, over 7346.00 frames. ], batch size: 61, lr: 5.96e-03, grad_scale: 8.0 +2023-03-21 04:34:40,232 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 04:34:48,188 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7886, 2.0117, 1.5422, 1.8591, 2.0284, 1.6495, 1.4176, 1.5928], + device='cuda:1'), covar=tensor([0.0135, 0.0223, 0.0323, 0.0186, 0.0104, 0.0237, 0.0332, 0.0128], + device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0029, 0.0029, 0.0030, 0.0028, 0.0028, 0.0030, 0.0038], + device='cuda:1'), out_proj_covar=tensor([3.6160e-05, 3.3013e-05, 3.2865e-05, 3.3256e-05, 3.2287e-05, 3.1621e-05, + 3.4396e-05, 4.3169e-05], device='cuda:1') +2023-03-21 04:35:06,111 INFO [train.py:935] (1/2) Epoch 28, validation: loss=0.1647, simple_loss=0.254, pruned_loss=0.03776, over 1622729.00 frames. +2023-03-21 04:35:06,112 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 04:35:13,303 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 04:35:16,485 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76269.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 04:35:23,488 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 04:35:24,614 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7861, 2.1294, 2.2579, 1.9203, 2.3315, 1.9434, 1.7120, 1.6856], + device='cuda:1'), covar=tensor([0.0816, 0.0402, 0.0389, 0.0245, 0.0257, 0.0567, 0.0319, 0.0371], + device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0031, 0.0033, 0.0030, 0.0031, 0.0031, 0.0034, 0.0034], + device='cuda:1'), out_proj_covar=tensor([8.4214e-05, 8.1960e-05, 8.1258e-05, 7.7354e-05, 8.0981e-05, 7.9327e-05, + 8.5046e-05, 8.6176e-05], device='cuda:1') +2023-03-21 04:35:29,098 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76294.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:35:30,409 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 04:35:31,421 INFO [train.py:901] (1/2) Epoch 28, batch 50, loss[loss=0.1065, simple_loss=0.1747, pruned_loss=0.01917, over 7096.00 frames. ], tot_loss[loss=0.1375, simple_loss=0.2177, pruned_loss=0.02867, over 324911.18 frames. ], batch size: 35, lr: 5.96e-03, grad_scale: 8.0 +2023-03-21 04:35:32,433 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 04:35:34,878 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 04:35:44,332 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76323.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 04:35:48,432 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76330.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 04:35:49,245 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.198e+02 1.914e+02 2.137e+02 2.399e+02 6.526e+02, threshold=4.275e+02, percent-clipped=3.0 +2023-03-21 04:35:52,771 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 04:35:53,295 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 04:35:56,209 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-03-21 04:35:57,765 INFO [train.py:901] (1/2) Epoch 28, batch 100, loss[loss=0.1425, simple_loss=0.2259, pruned_loss=0.02953, over 7298.00 frames. ], tot_loss[loss=0.1392, simple_loss=0.2196, pruned_loss=0.0294, over 572654.16 frames. ], batch size: 68, lr: 5.95e-03, grad_scale: 8.0 +2023-03-21 04:36:01,002 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76355.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:36:15,471 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76384.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 04:36:23,380 INFO [train.py:901] (1/2) Epoch 28, batch 150, loss[loss=0.1377, simple_loss=0.2252, pruned_loss=0.0251, over 7216.00 frames. ], tot_loss[loss=0.139, simple_loss=0.2195, pruned_loss=0.02927, over 765047.80 frames. ], batch size: 93, lr: 5.95e-03, grad_scale: 8.0 +2023-03-21 04:36:29,559 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-03-21 04:36:40,909 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.678e+02 1.996e+02 2.465e+02 6.288e+02, threshold=3.991e+02, percent-clipped=1.0 +2023-03-21 04:36:49,558 INFO [train.py:901] (1/2) Epoch 28, batch 200, loss[loss=0.159, simple_loss=0.2427, pruned_loss=0.03768, over 7134.00 frames. ], tot_loss[loss=0.1386, simple_loss=0.2192, pruned_loss=0.02901, over 916471.59 frames. ], batch size: 98, lr: 5.95e-03, grad_scale: 8.0 +2023-03-21 04:36:53,628 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 04:36:57,517 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 04:37:03,686 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 04:37:12,467 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6449, 3.3899, 2.3135, 3.8755, 3.1542, 3.4683, 1.8161, 2.2713], + device='cuda:1'), covar=tensor([0.0413, 0.0655, 0.2595, 0.0518, 0.0409, 0.0802, 0.3148, 0.1926], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0252, 0.0290, 0.0270, 0.0272, 0.0267, 0.0248, 0.0271], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:37:15,162 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 04:37:15,268 INFO [train.py:901] (1/2) Epoch 28, batch 250, loss[loss=0.1373, simple_loss=0.2139, pruned_loss=0.03036, over 7276.00 frames. ], tot_loss[loss=0.1393, simple_loss=0.2196, pruned_loss=0.0295, over 1033336.04 frames. ], batch size: 47, lr: 5.95e-03, grad_scale: 8.0 +2023-03-21 04:37:17,946 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 04:37:21,212 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 +2023-03-21 04:37:28,015 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1311, 3.8227, 3.8079, 3.8386, 3.7859, 3.7081, 4.0425, 3.6074], + device='cuda:1'), covar=tensor([0.0161, 0.0174, 0.0132, 0.0157, 0.0417, 0.0129, 0.0159, 0.0176], + device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0095, 0.0093, 0.0082, 0.0162, 0.0102, 0.0096, 0.0103], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:37:30,080 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8703, 4.3383, 4.3716, 4.2992, 4.4096, 3.8797, 4.3994, 4.3141], + device='cuda:1'), covar=tensor([0.0491, 0.0455, 0.0423, 0.0519, 0.0296, 0.0464, 0.0377, 0.0434], + device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0237, 0.0182, 0.0183, 0.0144, 0.0216, 0.0191, 0.0139], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 04:37:32,546 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.451e+02 1.788e+02 2.073e+02 2.472e+02 4.824e+02, threshold=4.147e+02, percent-clipped=3.0 +2023-03-21 04:37:37,677 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 04:37:41,152 INFO [train.py:901] (1/2) Epoch 28, batch 300, loss[loss=0.1371, simple_loss=0.2241, pruned_loss=0.02502, over 7317.00 frames. ], tot_loss[loss=0.1396, simple_loss=0.2198, pruned_loss=0.0297, over 1125218.62 frames. ], batch size: 80, lr: 5.95e-03, grad_scale: 8.0 +2023-03-21 04:37:46,251 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 04:37:58,023 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6036, 2.8476, 3.4260, 3.5427, 3.5760, 3.5999, 3.4021, 3.4510], + device='cuda:1'), covar=tensor([0.0028, 0.0122, 0.0038, 0.0036, 0.0035, 0.0030, 0.0065, 0.0051], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0064, 0.0054, 0.0052, 0.0051, 0.0056, 0.0047, 0.0070], + device='cuda:1'), out_proj_covar=tensor([8.2623e-05, 1.3940e-04, 1.0885e-04, 9.8807e-05, 9.4825e-05, 1.0479e-04, + 9.9139e-05, 1.3793e-04], device='cuda:1') +2023-03-21 04:38:07,546 INFO [train.py:901] (1/2) Epoch 28, batch 350, loss[loss=0.1367, simple_loss=0.2133, pruned_loss=0.03004, over 7269.00 frames. ], tot_loss[loss=0.1391, simple_loss=0.2191, pruned_loss=0.02948, over 1197497.03 frames. ], batch size: 47, lr: 5.94e-03, grad_scale: 8.0 +2023-03-21 04:38:21,269 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76625.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 04:38:21,311 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9782, 4.0978, 3.2787, 3.5752, 3.1556, 2.0635, 1.6471, 4.0444], + device='cuda:1'), covar=tensor([0.0069, 0.0060, 0.0189, 0.0085, 0.0200, 0.0698, 0.0788, 0.0077], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0082, 0.0103, 0.0089, 0.0117, 0.0124, 0.0123, 0.0096], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 04:38:23,194 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 04:38:24,663 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.636e+02 1.963e+02 2.462e+02 4.673e+02, threshold=3.925e+02, percent-clipped=2.0 +2023-03-21 04:38:25,293 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8762, 4.0881, 3.8406, 4.0865, 3.7223, 4.0531, 4.3093, 4.3835], + device='cuda:1'), covar=tensor([0.0234, 0.0157, 0.0196, 0.0149, 0.0359, 0.0260, 0.0243, 0.0167], + device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0119, 0.0108, 0.0116, 0.0106, 0.0095, 0.0097, 0.0092], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:38:33,375 INFO [train.py:901] (1/2) Epoch 28, batch 400, loss[loss=0.1197, simple_loss=0.2041, pruned_loss=0.01767, over 7257.00 frames. ], tot_loss[loss=0.1386, simple_loss=0.2188, pruned_loss=0.02924, over 1251646.77 frames. ], batch size: 89, lr: 5.94e-03, grad_scale: 8.0 +2023-03-21 04:38:33,971 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76650.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:38:48,298 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-03-21 04:38:49,646 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76679.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 04:38:59,617 INFO [train.py:901] (1/2) Epoch 28, batch 450, loss[loss=0.1244, simple_loss=0.1863, pruned_loss=0.03128, over 6102.00 frames. ], tot_loss[loss=0.1393, simple_loss=0.2196, pruned_loss=0.02955, over 1293160.28 frames. ], batch size: 26, lr: 5.94e-03, grad_scale: 8.0 +2023-03-21 04:39:03,599 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 04:39:04,118 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 04:39:16,076 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.221e+02 1.885e+02 2.115e+02 2.485e+02 5.641e+02, threshold=4.229e+02, percent-clipped=2.0 +2023-03-21 04:39:25,347 INFO [train.py:901] (1/2) Epoch 28, batch 500, loss[loss=0.1385, simple_loss=0.2205, pruned_loss=0.02824, over 7336.00 frames. ], tot_loss[loss=0.1386, simple_loss=0.2185, pruned_loss=0.02934, over 1322068.42 frames. ], batch size: 61, lr: 5.94e-03, grad_scale: 8.0 +2023-03-21 04:39:36,625 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 04:39:38,119 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 04:39:39,069 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 04:39:41,590 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 04:39:42,890 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 +2023-03-21 04:39:46,040 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 04:39:47,686 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9963, 3.9859, 3.3909, 3.5162, 2.9499, 2.2978, 1.8978, 3.9783], + device='cuda:1'), covar=tensor([0.0047, 0.0037, 0.0112, 0.0058, 0.0144, 0.0447, 0.0548, 0.0050], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0083, 0.0104, 0.0089, 0.0118, 0.0126, 0.0124, 0.0097], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 04:39:51,013 INFO [train.py:901] (1/2) Epoch 28, batch 550, loss[loss=0.1431, simple_loss=0.2285, pruned_loss=0.02886, over 7286.00 frames. ], tot_loss[loss=0.1388, simple_loss=0.2185, pruned_loss=0.02959, over 1348906.30 frames. ], batch size: 68, lr: 5.94e-03, grad_scale: 8.0 +2023-03-21 04:39:58,436 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 04:40:07,167 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 04:40:08,676 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.405e+02 1.821e+02 2.250e+02 2.684e+02 4.961e+02, threshold=4.501e+02, percent-clipped=3.0 +2023-03-21 04:40:09,738 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 04:40:11,406 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2127, 2.7242, 1.9847, 2.8803, 3.2674, 2.8290, 2.7551, 2.5584], + device='cuda:1'), covar=tensor([0.1741, 0.0822, 0.3433, 0.0444, 0.0195, 0.0164, 0.0266, 0.0304], + device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0229, 0.0257, 0.0262, 0.0184, 0.0181, 0.0204, 0.0220], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:40:12,351 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76839.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:40:16,357 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9174, 2.6475, 2.5322, 1.9706, 2.3672, 2.0884, 1.9109, 1.8325], + device='cuda:1'), covar=tensor([0.0513, 0.0303, 0.0173, 0.0310, 0.0602, 0.0587, 0.0278, 0.0412], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0031, 0.0033, 0.0030, 0.0031, 0.0031, 0.0035, 0.0034], + device='cuda:1'), out_proj_covar=tensor([8.5250e-05, 8.2041e-05, 8.1885e-05, 7.8111e-05, 8.1771e-05, 7.9984e-05, + 8.6202e-05, 8.6617e-05], device='cuda:1') +2023-03-21 04:40:16,770 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 04:40:17,289 INFO [train.py:901] (1/2) Epoch 28, batch 600, loss[loss=0.1371, simple_loss=0.2191, pruned_loss=0.02759, over 7254.00 frames. ], tot_loss[loss=0.1391, simple_loss=0.2185, pruned_loss=0.02983, over 1368816.59 frames. ], batch size: 89, lr: 5.93e-03, grad_scale: 8.0 +2023-03-21 04:40:22,451 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5114, 4.2463, 4.3626, 4.4920, 4.4697, 4.4680, 4.4994, 4.3401], + device='cuda:1'), covar=tensor([0.0020, 0.0058, 0.0033, 0.0027, 0.0023, 0.0028, 0.0025, 0.0043], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0062, 0.0053, 0.0051, 0.0050, 0.0055, 0.0047, 0.0069], + device='cuda:1'), out_proj_covar=tensor([8.0475e-05, 1.3600e-04, 1.0693e-04, 9.6122e-05, 9.1528e-05, 1.0290e-04, + 9.7995e-05, 1.3576e-04], device='cuda:1') +2023-03-21 04:40:25,385 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 +2023-03-21 04:40:33,032 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 04:40:42,040 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 04:40:42,989 INFO [train.py:901] (1/2) Epoch 28, batch 650, loss[loss=0.131, simple_loss=0.2108, pruned_loss=0.02561, over 7229.00 frames. ], tot_loss[loss=0.1386, simple_loss=0.2182, pruned_loss=0.0295, over 1383721.53 frames. ], batch size: 45, lr: 5.93e-03, grad_scale: 8.0 +2023-03-21 04:40:43,686 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76900.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:40:56,927 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76925.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 04:41:00,281 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.372e+02 1.758e+02 2.082e+02 2.489e+02 4.937e+02, threshold=4.165e+02, percent-clipped=1.0 +2023-03-21 04:41:00,329 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 04:41:09,284 INFO [train.py:901] (1/2) Epoch 28, batch 700, loss[loss=0.09765, simple_loss=0.1627, pruned_loss=0.01633, over 6050.00 frames. ], tot_loss[loss=0.1383, simple_loss=0.2181, pruned_loss=0.02928, over 1396422.20 frames. ], batch size: 26, lr: 5.93e-03, grad_scale: 8.0 +2023-03-21 04:41:09,312 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 04:41:09,894 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76950.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:41:21,251 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=76973.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 04:41:24,295 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76979.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 04:41:32,746 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 04:41:33,245 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 04:41:33,766 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=76998.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:41:34,206 INFO [train.py:901] (1/2) Epoch 28, batch 750, loss[loss=0.1279, simple_loss=0.2144, pruned_loss=0.02074, over 7331.00 frames. ], tot_loss[loss=0.1383, simple_loss=0.2178, pruned_loss=0.02933, over 1404770.75 frames. ], batch size: 75, lr: 5.93e-03, grad_scale: 16.0 +2023-03-21 04:41:47,123 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 04:41:49,146 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=77027.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 04:41:51,564 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.364e+02 1.849e+02 2.159e+02 2.501e+02 3.591e+02, threshold=4.317e+02, percent-clipped=0.0 +2023-03-21 04:41:52,117 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 04:41:58,777 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 04:42:00,307 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 04:42:00,789 INFO [train.py:901] (1/2) Epoch 28, batch 800, loss[loss=0.1429, simple_loss=0.2273, pruned_loss=0.02927, over 7294.00 frames. ], tot_loss[loss=0.1383, simple_loss=0.218, pruned_loss=0.02928, over 1415461.16 frames. ], batch size: 86, lr: 5.93e-03, grad_scale: 16.0 +2023-03-21 04:42:11,386 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 04:42:24,677 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77095.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:42:26,247 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.10 vs. limit=5.0 +2023-03-21 04:42:26,488 INFO [train.py:901] (1/2) Epoch 28, batch 850, loss[loss=0.1076, simple_loss=0.1793, pruned_loss=0.01796, over 6987.00 frames. ], tot_loss[loss=0.1378, simple_loss=0.2176, pruned_loss=0.029, over 1422164.01 frames. ], batch size: 35, lr: 5.92e-03, grad_scale: 4.0 +2023-03-21 04:42:27,654 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3539, 2.9084, 2.1489, 3.1513, 3.4765, 3.2592, 2.7567, 2.6457], + device='cuda:1'), covar=tensor([0.1782, 0.0797, 0.3496, 0.0384, 0.0200, 0.0182, 0.0391, 0.0337], + device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0231, 0.0258, 0.0262, 0.0184, 0.0182, 0.0206, 0.0222], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 04:42:30,924 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 04:42:30,932 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 04:42:36,551 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 04:42:40,670 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 04:42:44,711 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.361e+02 1.829e+02 2.089e+02 2.425e+02 3.800e+02, threshold=4.177e+02, percent-clipped=0.0 +2023-03-21 04:42:52,367 INFO [train.py:901] (1/2) Epoch 28, batch 900, loss[loss=0.128, simple_loss=0.2196, pruned_loss=0.01818, over 7348.00 frames. ], tot_loss[loss=0.1382, simple_loss=0.2177, pruned_loss=0.02932, over 1423361.89 frames. ], batch size: 63, lr: 5.92e-03, grad_scale: 4.0 +2023-03-21 04:42:56,099 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77156.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 04:43:05,042 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8647, 3.5936, 3.5957, 3.5792, 3.4740, 3.4602, 3.7189, 3.4051], + device='cuda:1'), covar=tensor([0.0135, 0.0180, 0.0137, 0.0175, 0.0432, 0.0122, 0.0176, 0.0160], + device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0094, 0.0092, 0.0082, 0.0161, 0.0101, 0.0096, 0.0102], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:43:15,984 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77195.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:43:17,921 INFO [train.py:901] (1/2) Epoch 28, batch 950, loss[loss=0.1353, simple_loss=0.2196, pruned_loss=0.02553, over 7323.00 frames. ], tot_loss[loss=0.1397, simple_loss=0.219, pruned_loss=0.03017, over 1425400.94 frames. ], batch size: 75, lr: 5.92e-03, grad_scale: 4.0 +2023-03-21 04:43:18,409 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 04:43:21,580 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4948, 2.4516, 2.2472, 3.6601, 1.6415, 3.4847, 1.4039, 3.0094], + device='cuda:1'), covar=tensor([0.0169, 0.1157, 0.1646, 0.0145, 0.3954, 0.0192, 0.1277, 0.0305], + device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0257, 0.0269, 0.0196, 0.0258, 0.0207, 0.0242, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:43:25,102 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77211.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:43:25,321 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.08 vs. limit=5.0 +2023-03-21 04:43:36,542 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.269e+02 1.862e+02 2.179e+02 2.499e+02 5.050e+02, threshold=4.359e+02, percent-clipped=3.0 +2023-03-21 04:43:38,172 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2768, 4.7527, 4.8273, 4.7087, 4.6692, 4.2469, 4.8652, 4.6452], + device='cuda:1'), covar=tensor([0.0457, 0.0378, 0.0335, 0.0565, 0.0347, 0.0419, 0.0283, 0.0455], + device='cuda:1'), in_proj_covar=tensor([0.0211, 0.0243, 0.0186, 0.0190, 0.0147, 0.0220, 0.0195, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 04:43:42,204 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 04:43:44,219 INFO [train.py:901] (1/2) Epoch 28, batch 1000, loss[loss=0.1358, simple_loss=0.2103, pruned_loss=0.03059, over 7264.00 frames. ], tot_loss[loss=0.1395, simple_loss=0.2191, pruned_loss=0.03, over 1428989.76 frames. ], batch size: 47, lr: 5.92e-03, grad_scale: 4.0 +2023-03-21 04:43:55,414 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-21 04:43:56,706 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77272.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:43:57,159 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77273.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:43:57,196 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9289, 2.4293, 1.8691, 2.5965, 2.7742, 2.6164, 2.4914, 2.4020], + device='cuda:1'), covar=tensor([0.1894, 0.0915, 0.3162, 0.0566, 0.0212, 0.0247, 0.0335, 0.0339], + device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0233, 0.0258, 0.0265, 0.0187, 0.0185, 0.0209, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 04:43:59,155 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77277.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:43:59,177 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9165, 2.6969, 2.5039, 3.9667, 1.7542, 3.7763, 1.3936, 3.3101], + device='cuda:1'), covar=tensor([0.0135, 0.1056, 0.1532, 0.0159, 0.3984, 0.0186, 0.1161, 0.0370], + device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0258, 0.0273, 0.0197, 0.0259, 0.0208, 0.0244, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:44:02,520 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 04:44:10,570 INFO [train.py:901] (1/2) Epoch 28, batch 1050, loss[loss=0.1373, simple_loss=0.2198, pruned_loss=0.02739, over 7309.00 frames. ], tot_loss[loss=0.1396, simple_loss=0.2195, pruned_loss=0.02981, over 1434346.13 frames. ], batch size: 86, lr: 5.92e-03, grad_scale: 4.0 +2023-03-21 04:44:24,252 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 04:44:28,250 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.790e+02 2.098e+02 2.468e+02 4.594e+02, threshold=4.197e+02, percent-clipped=2.0 +2023-03-21 04:44:28,304 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 04:44:28,404 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77334.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:44:30,398 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77338.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:44:35,855 INFO [train.py:901] (1/2) Epoch 28, batch 1100, loss[loss=0.1348, simple_loss=0.2129, pruned_loss=0.02834, over 7242.00 frames. ], tot_loss[loss=0.1392, simple_loss=0.2193, pruned_loss=0.02961, over 1437149.63 frames. ], batch size: 55, lr: 5.91e-03, grad_scale: 4.0 +2023-03-21 04:44:46,154 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7188, 3.1641, 2.4969, 3.6964, 3.6259, 3.4443, 3.5689, 3.3552], + device='cuda:1'), covar=tensor([0.1493, 0.0531, 0.2831, 0.0557, 0.0137, 0.0146, 0.0252, 0.0299], + device='cuda:1'), in_proj_covar=tensor([0.0261, 0.0232, 0.0258, 0.0265, 0.0186, 0.0184, 0.0210, 0.0223], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 04:44:56,719 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1957, 2.8492, 3.1896, 3.0471, 2.7066, 2.7363, 2.8522, 2.3753], + device='cuda:1'), covar=tensor([0.0443, 0.0454, 0.0486, 0.0537, 0.0599, 0.0790, 0.0536, 0.1683], + device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0343, 0.0274, 0.0362, 0.0300, 0.0298, 0.0348, 0.0272], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:44:57,065 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 04:44:57,515 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 04:45:01,308 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.91 vs. limit=5.0 +2023-03-21 04:45:01,972 INFO [train.py:901] (1/2) Epoch 28, batch 1150, loss[loss=0.1628, simple_loss=0.2458, pruned_loss=0.03986, over 7237.00 frames. ], tot_loss[loss=0.1396, simple_loss=0.2198, pruned_loss=0.02967, over 1438704.22 frames. ], batch size: 64, lr: 5.91e-03, grad_scale: 4.0 +2023-03-21 04:45:04,088 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.81 vs. limit=5.0 +2023-03-21 04:45:10,361 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 04:45:10,878 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 04:45:19,868 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.393e+02 1.883e+02 2.182e+02 2.626e+02 3.951e+02, threshold=4.364e+02, percent-clipped=0.0 +2023-03-21 04:45:27,935 INFO [train.py:901] (1/2) Epoch 28, batch 1200, loss[loss=0.1579, simple_loss=0.2437, pruned_loss=0.03603, over 6692.00 frames. ], tot_loss[loss=0.1405, simple_loss=0.2204, pruned_loss=0.03036, over 1439265.46 frames. ], batch size: 106, lr: 5.91e-03, grad_scale: 8.0 +2023-03-21 04:45:28,992 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77451.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 04:45:41,089 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0691, 2.5717, 1.9070, 2.9578, 2.9429, 2.7492, 2.6649, 2.5780], + device='cuda:1'), covar=tensor([0.2101, 0.0922, 0.3504, 0.0621, 0.0193, 0.0171, 0.0349, 0.0289], + device='cuda:1'), in_proj_covar=tensor([0.0261, 0.0232, 0.0259, 0.0264, 0.0186, 0.0184, 0.0208, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 04:45:43,450 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 04:45:51,513 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77495.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:45:53,438 INFO [train.py:901] (1/2) Epoch 28, batch 1250, loss[loss=0.1421, simple_loss=0.2292, pruned_loss=0.02746, over 7295.00 frames. ], tot_loss[loss=0.1398, simple_loss=0.2196, pruned_loss=0.02999, over 1440585.44 frames. ], batch size: 86, lr: 5.91e-03, grad_scale: 8.0 +2023-03-21 04:46:00,780 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-03-21 04:46:06,111 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 04:46:10,733 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 04:46:11,740 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.879e+02 2.180e+02 2.544e+02 4.448e+02, threshold=4.361e+02, percent-clipped=1.0 +2023-03-21 04:46:11,789 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 04:46:16,348 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=77543.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:46:19,406 INFO [train.py:901] (1/2) Epoch 28, batch 1300, loss[loss=0.1471, simple_loss=0.232, pruned_loss=0.03114, over 7255.00 frames. ], tot_loss[loss=0.1393, simple_loss=0.2192, pruned_loss=0.02974, over 1441245.76 frames. ], batch size: 52, lr: 5.91e-03, grad_scale: 8.0 +2023-03-21 04:46:29,040 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77567.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:46:35,863 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-03-21 04:46:36,039 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 04:46:38,015 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 04:46:41,494 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 04:46:44,997 INFO [train.py:901] (1/2) Epoch 28, batch 1350, loss[loss=0.1493, simple_loss=0.2243, pruned_loss=0.03713, over 7306.00 frames. ], tot_loss[loss=0.139, simple_loss=0.2188, pruned_loss=0.02956, over 1441375.98 frames. ], batch size: 86, lr: 5.91e-03, grad_scale: 8.0 +2023-03-21 04:46:46,944 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9038, 2.3671, 3.0724, 2.9601, 3.1352, 2.7074, 2.4158, 3.0475], + device='cuda:1'), covar=tensor([0.1469, 0.0856, 0.0906, 0.1152, 0.0578, 0.1365, 0.2277, 0.1098], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0064, 0.0049, 0.0047, 0.0046, 0.0046, 0.0065, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 04:46:51,739 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 04:46:54,571 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77616.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:47:01,088 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77629.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:47:03,183 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77633.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:47:03,584 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.276e+02 1.842e+02 2.121e+02 2.615e+02 6.191e+02, threshold=4.241e+02, percent-clipped=3.0 +2023-03-21 04:47:11,938 INFO [train.py:901] (1/2) Epoch 28, batch 1400, loss[loss=0.1162, simple_loss=0.1907, pruned_loss=0.02081, over 7194.00 frames. ], tot_loss[loss=0.1388, simple_loss=0.2188, pruned_loss=0.02942, over 1442708.38 frames. ], batch size: 39, lr: 5.90e-03, grad_scale: 8.0 +2023-03-21 04:47:17,537 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9567, 2.9721, 3.4333, 3.3059, 3.2743, 3.1414, 2.5993, 3.3117], + device='cuda:1'), covar=tensor([0.2337, 0.0759, 0.1163, 0.1081, 0.0780, 0.1134, 0.2559, 0.1654], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0065, 0.0049, 0.0048, 0.0047, 0.0046, 0.0066, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 04:47:26,056 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 04:47:26,146 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:47:26,159 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:47:37,269 INFO [train.py:901] (1/2) Epoch 28, batch 1450, loss[loss=0.1497, simple_loss=0.234, pruned_loss=0.03271, over 7285.00 frames. ], tot_loss[loss=0.1385, simple_loss=0.2184, pruned_loss=0.02927, over 1440953.52 frames. ], batch size: 77, lr: 5.90e-03, grad_scale: 8.0 +2023-03-21 04:47:45,587 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9342, 2.8300, 2.9468, 3.0976, 2.6668, 2.7197, 2.8791, 2.3868], + device='cuda:1'), covar=tensor([0.0443, 0.0532, 0.0515, 0.0527, 0.0536, 0.0797, 0.0643, 0.1509], + device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0341, 0.0271, 0.0360, 0.0297, 0.0298, 0.0345, 0.0270], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:47:50,527 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 04:47:55,968 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.062e+02 1.839e+02 2.145e+02 2.638e+02 3.761e+02, threshold=4.290e+02, percent-clipped=0.0 +2023-03-21 04:47:58,166 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77738.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:48:03,456 INFO [train.py:901] (1/2) Epoch 28, batch 1500, loss[loss=0.1496, simple_loss=0.2285, pruned_loss=0.03532, over 7259.00 frames. ], tot_loss[loss=0.1389, simple_loss=0.219, pruned_loss=0.0294, over 1442112.60 frames. ], batch size: 47, lr: 5.90e-03, grad_scale: 8.0 +2023-03-21 04:48:04,585 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77751.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 04:48:07,587 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 04:48:26,769 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-03-21 04:48:29,475 INFO [train.py:901] (1/2) Epoch 28, batch 1550, loss[loss=0.1243, simple_loss=0.2029, pruned_loss=0.02279, over 7149.00 frames. ], tot_loss[loss=0.1385, simple_loss=0.2185, pruned_loss=0.02924, over 1440378.65 frames. ], batch size: 41, lr: 5.90e-03, grad_scale: 8.0 +2023-03-21 04:48:29,545 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=77799.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:48:32,336 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 04:48:48,025 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.264e+02 1.872e+02 2.235e+02 2.567e+02 5.065e+02, threshold=4.469e+02, percent-clipped=1.0 +2023-03-21 04:48:55,576 INFO [train.py:901] (1/2) Epoch 28, batch 1600, loss[loss=0.1146, simple_loss=0.1993, pruned_loss=0.01493, over 7317.00 frames. ], tot_loss[loss=0.1386, simple_loss=0.2184, pruned_loss=0.02938, over 1440796.96 frames. ], batch size: 44, lr: 5.90e-03, grad_scale: 8.0 +2023-03-21 04:49:03,072 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 04:49:04,051 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 04:49:04,659 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77867.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:49:05,344 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 +2023-03-21 04:49:06,589 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 04:49:16,773 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 04:49:21,315 INFO [train.py:901] (1/2) Epoch 28, batch 1650, loss[loss=0.1498, simple_loss=0.2306, pruned_loss=0.03449, over 7278.00 frames. ], tot_loss[loss=0.1393, simple_loss=0.2192, pruned_loss=0.02972, over 1440575.66 frames. ], batch size: 77, lr: 5.89e-03, grad_scale: 8.0 +2023-03-21 04:49:21,332 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 04:49:23,059 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5438, 3.3715, 2.5345, 3.8749, 3.0076, 3.5596, 1.8793, 2.4731], + device='cuda:1'), covar=tensor([0.0379, 0.0941, 0.2225, 0.0457, 0.0421, 0.0656, 0.3183, 0.1937], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0251, 0.0289, 0.0266, 0.0268, 0.0265, 0.0245, 0.0266], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:49:29,483 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 04:49:29,973 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=77915.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:49:37,095 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77929.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:49:39,072 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77933.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:49:39,433 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.287e+02 1.899e+02 2.148e+02 2.752e+02 4.294e+02, threshold=4.295e+02, percent-clipped=0.0 +2023-03-21 04:49:46,490 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 04:49:46,979 INFO [train.py:901] (1/2) Epoch 28, batch 1700, loss[loss=0.1305, simple_loss=0.2163, pruned_loss=0.0224, over 7264.00 frames. ], tot_loss[loss=0.1394, simple_loss=0.2192, pruned_loss=0.02976, over 1441626.25 frames. ], batch size: 64, lr: 5.89e-03, grad_scale: 8.0 +2023-03-21 04:49:50,019 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 04:49:59,395 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77972.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:50:01,341 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 04:50:01,919 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=77977.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:50:03,953 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=77981.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:50:13,650 INFO [train.py:901] (1/2) Epoch 28, batch 1750, loss[loss=0.1342, simple_loss=0.2167, pruned_loss=0.02586, over 7284.00 frames. ], tot_loss[loss=0.1399, simple_loss=0.2196, pruned_loss=0.03008, over 1443399.41 frames. ], batch size: 70, lr: 5.89e-03, grad_scale: 8.0 +2023-03-21 04:50:26,529 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 04:50:28,103 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 04:50:31,244 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78033.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:50:31,656 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.311e+02 1.834e+02 2.126e+02 2.360e+02 4.852e+02, threshold=4.252e+02, percent-clipped=1.0 +2023-03-21 04:50:36,265 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0330, 2.2760, 2.5274, 2.0500, 2.1077, 2.3399, 2.0752, 1.9194], + device='cuda:1'), covar=tensor([0.0469, 0.0402, 0.0323, 0.0224, 0.0714, 0.0353, 0.0317, 0.0458], + device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0032, 0.0032, 0.0030, 0.0030, 0.0030, 0.0035, 0.0034], + device='cuda:1'), out_proj_covar=tensor([8.4176e-05, 8.2566e-05, 8.0539e-05, 7.7793e-05, 8.0012e-05, 7.8821e-05, + 8.5539e-05, 8.6320e-05], device='cuda:1') +2023-03-21 04:50:36,434 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 +2023-03-21 04:50:39,128 INFO [train.py:901] (1/2) Epoch 28, batch 1800, loss[loss=0.1534, simple_loss=0.2359, pruned_loss=0.03541, over 7256.00 frames. ], tot_loss[loss=0.1393, simple_loss=0.2188, pruned_loss=0.02983, over 1441926.27 frames. ], batch size: 55, lr: 5.89e-03, grad_scale: 8.0 +2023-03-21 04:50:39,510 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 +2023-03-21 04:50:50,127 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.58 vs. limit=5.0 +2023-03-21 04:50:50,380 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 04:50:52,169 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.36 vs. limit=5.0 +2023-03-21 04:50:53,000 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1034, 2.6651, 1.9576, 2.9653, 2.7528, 2.8511, 2.5870, 2.3568], + device='cuda:1'), covar=tensor([0.2015, 0.0810, 0.3284, 0.0488, 0.0177, 0.0172, 0.0302, 0.0364], + device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0232, 0.0259, 0.0263, 0.0187, 0.0184, 0.0207, 0.0223], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 04:51:04,990 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 04:51:05,475 INFO [train.py:901] (1/2) Epoch 28, batch 1850, loss[loss=0.1479, simple_loss=0.2267, pruned_loss=0.03453, over 7315.00 frames. ], tot_loss[loss=0.1396, simple_loss=0.2193, pruned_loss=0.02993, over 1440328.49 frames. ], batch size: 49, lr: 5.89e-03, grad_scale: 8.0 +2023-03-21 04:51:14,556 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 04:51:23,152 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.319e+02 1.957e+02 2.174e+02 2.564e+02 5.105e+02, threshold=4.348e+02, percent-clipped=1.0 +2023-03-21 04:51:27,822 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0190, 3.6349, 3.9676, 3.9891, 4.0256, 4.0485, 4.2441, 3.9409], + device='cuda:1'), covar=tensor([0.0029, 0.0084, 0.0031, 0.0031, 0.0032, 0.0031, 0.0028, 0.0051], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0062, 0.0052, 0.0051, 0.0050, 0.0054, 0.0046, 0.0068], + device='cuda:1'), out_proj_covar=tensor([7.9990e-05, 1.3498e-04, 1.0350e-04, 9.5183e-05, 9.0445e-05, 1.0152e-04, + 9.6061e-05, 1.3334e-04], device='cuda:1') +2023-03-21 04:51:31,393 INFO [train.py:901] (1/2) Epoch 28, batch 1900, loss[loss=0.1362, simple_loss=0.2103, pruned_loss=0.03102, over 7217.00 frames. ], tot_loss[loss=0.1391, simple_loss=0.2186, pruned_loss=0.02977, over 1440912.85 frames. ], batch size: 45, lr: 5.88e-03, grad_scale: 8.0 +2023-03-21 04:51:31,887 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 04:51:49,729 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9950, 2.4820, 1.7865, 3.2654, 2.7956, 3.1306, 2.7674, 2.4466], + device='cuda:1'), covar=tensor([0.2205, 0.1054, 0.3708, 0.0566, 0.0225, 0.0233, 0.0321, 0.0336], + device='cuda:1'), in_proj_covar=tensor([0.0264, 0.0234, 0.0260, 0.0266, 0.0188, 0.0184, 0.0207, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 04:51:56,645 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 04:51:57,179 INFO [train.py:901] (1/2) Epoch 28, batch 1950, loss[loss=0.1448, simple_loss=0.2208, pruned_loss=0.03439, over 7350.00 frames. ], tot_loss[loss=0.1387, simple_loss=0.2183, pruned_loss=0.02959, over 1441567.61 frames. ], batch size: 51, lr: 5.88e-03, grad_scale: 8.0 +2023-03-21 04:52:01,144 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3298, 3.2986, 2.2327, 3.7126, 2.7781, 3.4080, 1.5644, 2.1583], + device='cuda:1'), covar=tensor([0.0403, 0.0987, 0.2585, 0.0565, 0.0500, 0.0697, 0.3367, 0.1913], + device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0255, 0.0292, 0.0272, 0.0272, 0.0268, 0.0250, 0.0271], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:52:06,932 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 04:52:12,108 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 04:52:12,620 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 04:52:14,697 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78232.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:52:15,521 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.183e+02 1.838e+02 2.132e+02 2.471e+02 5.256e+02, threshold=4.265e+02, percent-clipped=3.0 +2023-03-21 04:52:23,075 INFO [train.py:901] (1/2) Epoch 28, batch 2000, loss[loss=0.1448, simple_loss=0.2264, pruned_loss=0.03157, over 7282.00 frames. ], tot_loss[loss=0.1394, simple_loss=0.2191, pruned_loss=0.02984, over 1442817.70 frames. ], batch size: 52, lr: 5.88e-03, grad_scale: 8.0 +2023-03-21 04:52:30,191 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 04:52:35,228 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78272.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:52:40,599 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 04:52:45,743 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78293.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 04:52:48,085 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 04:52:48,577 INFO [train.py:901] (1/2) Epoch 28, batch 2050, loss[loss=0.1538, simple_loss=0.2289, pruned_loss=0.03935, over 7356.00 frames. ], tot_loss[loss=0.1401, simple_loss=0.2199, pruned_loss=0.03014, over 1443761.14 frames. ], batch size: 63, lr: 5.88e-03, grad_scale: 8.0 +2023-03-21 04:52:57,178 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-03-21 04:52:59,898 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=78320.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:53:00,534 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5098, 2.3272, 2.1995, 3.6491, 1.7884, 3.4549, 1.4810, 3.1231], + device='cuda:1'), covar=tensor([0.0129, 0.1256, 0.1813, 0.0147, 0.3680, 0.0197, 0.1108, 0.0347], + device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0262, 0.0275, 0.0199, 0.0263, 0.0213, 0.0247, 0.0233], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:53:06,533 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78333.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:53:06,910 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.335e+02 1.680e+02 2.089e+02 2.473e+02 3.796e+02, threshold=4.179e+02, percent-clipped=0.0 +2023-03-21 04:53:15,214 INFO [train.py:901] (1/2) Epoch 28, batch 2100, loss[loss=0.1419, simple_loss=0.2223, pruned_loss=0.03077, over 7355.00 frames. ], tot_loss[loss=0.1395, simple_loss=0.2193, pruned_loss=0.02988, over 1443901.91 frames. ], batch size: 63, lr: 5.88e-03, grad_scale: 8.0 +2023-03-21 04:53:16,836 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3774, 3.8847, 3.8169, 4.4024, 4.2937, 4.3364, 3.8133, 3.9139], + device='cuda:1'), covar=tensor([0.0930, 0.2729, 0.2558, 0.1216, 0.0977, 0.1299, 0.0968, 0.1335], + device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0361, 0.0279, 0.0287, 0.0210, 0.0347, 0.0206, 0.0255], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:53:18,412 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.60 vs. limit=5.0 +2023-03-21 04:53:21,708 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 04:53:24,152 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 04:53:31,129 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=78381.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:53:40,730 INFO [train.py:901] (1/2) Epoch 28, batch 2150, loss[loss=0.1409, simple_loss=0.225, pruned_loss=0.02837, over 7223.00 frames. ], tot_loss[loss=0.1398, simple_loss=0.2198, pruned_loss=0.02991, over 1443069.07 frames. ], batch size: 93, lr: 5.88e-03, grad_scale: 8.0 +2023-03-21 04:53:58,831 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.2663, 5.6845, 5.7796, 5.7208, 5.4612, 5.4164, 5.7544, 5.6190], + device='cuda:1'), covar=tensor([0.0409, 0.0363, 0.0250, 0.0344, 0.0255, 0.0296, 0.0310, 0.0300], + device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0240, 0.0183, 0.0185, 0.0146, 0.0218, 0.0193, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 04:53:59,252 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.886e+02 2.193e+02 2.710e+02 5.513e+02, threshold=4.386e+02, percent-clipped=4.0 +2023-03-21 04:54:06,878 INFO [train.py:901] (1/2) Epoch 28, batch 2200, loss[loss=0.1453, simple_loss=0.2307, pruned_loss=0.02998, over 7285.00 frames. ], tot_loss[loss=0.14, simple_loss=0.22, pruned_loss=0.02997, over 1440366.15 frames. ], batch size: 86, lr: 5.87e-03, grad_scale: 8.0 +2023-03-21 04:54:10,944 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 04:54:12,468 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78460.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:54:17,411 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9065, 2.9787, 2.7019, 2.9933, 3.0245, 2.5380, 3.0450, 2.6296], + device='cuda:1'), covar=tensor([0.0593, 0.1075, 0.1413, 0.1606, 0.1088, 0.0744, 0.1162, 0.1537], + device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0054, 0.0062, 0.0054, 0.0051, 0.0056, 0.0052, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:54:32,504 INFO [train.py:901] (1/2) Epoch 28, batch 2250, loss[loss=0.1159, simple_loss=0.1871, pruned_loss=0.02231, over 7011.00 frames. ], tot_loss[loss=0.1394, simple_loss=0.2196, pruned_loss=0.02957, over 1441210.69 frames. ], batch size: 35, lr: 5.87e-03, grad_scale: 8.0 +2023-03-21 04:54:41,142 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9720, 4.1435, 3.9129, 4.1220, 3.6294, 4.1086, 4.3574, 4.4265], + device='cuda:1'), covar=tensor([0.0189, 0.0166, 0.0195, 0.0150, 0.0442, 0.0243, 0.0230, 0.0149], + device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0120, 0.0109, 0.0115, 0.0108, 0.0097, 0.0095, 0.0093], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:54:44,206 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78521.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:54:45,108 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 04:54:45,119 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 04:54:50,362 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.188e+02 1.836e+02 2.134e+02 2.527e+02 4.591e+02, threshold=4.269e+02, percent-clipped=1.0 +2023-03-21 04:54:56,407 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 04:54:57,853 INFO [train.py:901] (1/2) Epoch 28, batch 2300, loss[loss=0.1185, simple_loss=0.1911, pruned_loss=0.02294, over 7201.00 frames. ], tot_loss[loss=0.1391, simple_loss=0.2194, pruned_loss=0.02943, over 1440879.02 frames. ], batch size: 39, lr: 5.87e-03, grad_scale: 8.0 +2023-03-21 04:55:01,385 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78556.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:55:17,917 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78588.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 04:55:23,535 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.9849, 4.4408, 4.3727, 4.8512, 4.8008, 4.8992, 4.4114, 4.5003], + device='cuda:1'), covar=tensor([0.0813, 0.2745, 0.2320, 0.1218, 0.0883, 0.1155, 0.0756, 0.1158], + device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0366, 0.0283, 0.0291, 0.0212, 0.0351, 0.0209, 0.0258], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:55:23,958 INFO [train.py:901] (1/2) Epoch 28, batch 2350, loss[loss=0.1495, simple_loss=0.2358, pruned_loss=0.0316, over 7354.00 frames. ], tot_loss[loss=0.1397, simple_loss=0.22, pruned_loss=0.02968, over 1442168.39 frames. ], batch size: 63, lr: 5.87e-03, grad_scale: 8.0 +2023-03-21 04:55:33,488 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78617.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:55:41,552 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 04:55:42,032 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.400e+02 1.841e+02 2.209e+02 2.542e+02 3.662e+02, threshold=4.418e+02, percent-clipped=0.0 +2023-03-21 04:55:43,237 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3085, 3.2099, 2.4658, 3.7727, 2.8542, 3.1556, 1.5412, 2.2539], + device='cuda:1'), covar=tensor([0.0396, 0.0925, 0.2199, 0.0496, 0.0500, 0.0708, 0.3353, 0.1905], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0255, 0.0289, 0.0273, 0.0274, 0.0269, 0.0248, 0.0271], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 04:55:47,553 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 04:55:50,111 INFO [train.py:901] (1/2) Epoch 28, batch 2400, loss[loss=0.1633, simple_loss=0.2407, pruned_loss=0.043, over 7273.00 frames. ], tot_loss[loss=0.1396, simple_loss=0.2199, pruned_loss=0.02965, over 1441874.14 frames. ], batch size: 64, lr: 5.87e-03, grad_scale: 8.0 +2023-03-21 04:55:59,261 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 04:56:01,830 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 04:56:15,914 INFO [train.py:901] (1/2) Epoch 28, batch 2450, loss[loss=0.1585, simple_loss=0.236, pruned_loss=0.0405, over 7349.00 frames. ], tot_loss[loss=0.1391, simple_loss=0.2194, pruned_loss=0.02937, over 1440526.30 frames. ], batch size: 63, lr: 5.86e-03, grad_scale: 8.0 +2023-03-21 04:56:17,544 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8673, 3.2401, 3.7286, 3.7587, 3.7531, 3.8520, 3.9152, 3.7477], + device='cuda:1'), covar=tensor([0.0031, 0.0105, 0.0038, 0.0034, 0.0038, 0.0031, 0.0033, 0.0053], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0064, 0.0055, 0.0053, 0.0052, 0.0056, 0.0048, 0.0071], + device='cuda:1'), out_proj_covar=tensor([8.2891e-05, 1.3967e-04, 1.0840e-04, 9.8428e-05, 9.4688e-05, 1.0479e-04, + 9.9867e-05, 1.3959e-04], device='cuda:1') +2023-03-21 04:56:29,117 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 04:56:33,612 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.829e+02 2.117e+02 2.491e+02 4.848e+02, threshold=4.235e+02, percent-clipped=1.0 +2023-03-21 04:56:41,768 INFO [train.py:901] (1/2) Epoch 28, batch 2500, loss[loss=0.1402, simple_loss=0.2179, pruned_loss=0.03127, over 7284.00 frames. ], tot_loss[loss=0.1387, simple_loss=0.2189, pruned_loss=0.02927, over 1438869.41 frames. ], batch size: 66, lr: 5.86e-03, grad_scale: 8.0 +2023-03-21 04:56:55,492 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 04:57:05,567 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.8441, 5.3968, 5.5032, 5.3971, 5.1528, 4.8803, 5.5142, 5.2604], + device='cuda:1'), covar=tensor([0.0367, 0.0316, 0.0273, 0.0378, 0.0290, 0.0325, 0.0260, 0.0471], + device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0240, 0.0183, 0.0186, 0.0146, 0.0218, 0.0193, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 04:57:07,548 INFO [train.py:901] (1/2) Epoch 28, batch 2550, loss[loss=0.1591, simple_loss=0.2322, pruned_loss=0.04303, over 7358.00 frames. ], tot_loss[loss=0.1381, simple_loss=0.218, pruned_loss=0.02905, over 1438231.36 frames. ], batch size: 73, lr: 5.86e-03, grad_scale: 8.0 +2023-03-21 04:57:16,614 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78816.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:57:26,232 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.209e+02 1.802e+02 2.097e+02 2.497e+02 4.089e+02, threshold=4.193e+02, percent-clipped=0.0 +2023-03-21 04:57:29,460 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0980, 2.0409, 2.2738, 1.8867, 2.2227, 2.1330, 1.8330, 1.6017], + device='cuda:1'), covar=tensor([0.0312, 0.0314, 0.0176, 0.0215, 0.0278, 0.0274, 0.0340, 0.0223], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0033, 0.0033, 0.0031, 0.0031, 0.0031, 0.0035, 0.0034], + device='cuda:1'), out_proj_covar=tensor([8.5942e-05, 8.4683e-05, 8.2275e-05, 7.9387e-05, 8.1383e-05, 8.0211e-05, + 8.6204e-05, 8.7763e-05], device='cuda:1') +2023-03-21 04:57:33,692 INFO [train.py:901] (1/2) Epoch 28, batch 2600, loss[loss=0.1437, simple_loss=0.2234, pruned_loss=0.03197, over 7209.00 frames. ], tot_loss[loss=0.1386, simple_loss=0.2185, pruned_loss=0.02938, over 1440984.97 frames. ], batch size: 45, lr: 5.86e-03, grad_scale: 8.0 +2023-03-21 04:57:51,960 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4754, 3.1261, 3.3273, 3.3691, 3.0639, 2.8984, 3.4530, 2.6522], + device='cuda:1'), covar=tensor([0.0378, 0.0406, 0.0522, 0.0594, 0.0569, 0.0775, 0.0558, 0.1557], + device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0340, 0.0271, 0.0359, 0.0300, 0.0297, 0.0344, 0.0270], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 04:57:52,438 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6600, 2.9279, 2.4923, 2.7420, 2.8788, 2.5530, 2.7852, 2.7240], + device='cuda:1'), covar=tensor([0.0691, 0.0741, 0.1670, 0.1912, 0.0940, 0.1035, 0.1067, 0.1041], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0054, 0.0061, 0.0054, 0.0051, 0.0056, 0.0052, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:57:52,877 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78888.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:57:58,178 INFO [train.py:901] (1/2) Epoch 28, batch 2650, loss[loss=0.151, simple_loss=0.2234, pruned_loss=0.03934, over 7292.00 frames. ], tot_loss[loss=0.1388, simple_loss=0.2185, pruned_loss=0.02951, over 1442037.08 frames. ], batch size: 86, lr: 5.86e-03, grad_scale: 8.0 +2023-03-21 04:58:04,643 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78912.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:58:15,876 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.313e+02 1.795e+02 2.054e+02 2.347e+02 4.847e+02, threshold=4.109e+02, percent-clipped=2.0 +2023-03-21 04:58:16,977 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=78936.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:58:19,972 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.3578, 4.8381, 4.6735, 5.3453, 5.1404, 5.2309, 4.7293, 4.8043], + device='cuda:1'), covar=tensor([0.0696, 0.2714, 0.2153, 0.0981, 0.0865, 0.1220, 0.0699, 0.1031], + device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0363, 0.0278, 0.0284, 0.0209, 0.0346, 0.0205, 0.0254], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:58:23,306 INFO [train.py:901] (1/2) Epoch 28, batch 2700, loss[loss=0.1372, simple_loss=0.2201, pruned_loss=0.02713, over 7279.00 frames. ], tot_loss[loss=0.1383, simple_loss=0.2183, pruned_loss=0.02916, over 1444656.82 frames. ], batch size: 68, lr: 5.85e-03, grad_scale: 8.0 +2023-03-21 04:58:31,343 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78965.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:58:47,900 INFO [train.py:901] (1/2) Epoch 28, batch 2750, loss[loss=0.131, simple_loss=0.2111, pruned_loss=0.02543, over 7255.00 frames. ], tot_loss[loss=0.1376, simple_loss=0.2177, pruned_loss=0.02875, over 1443613.05 frames. ], batch size: 52, lr: 5.85e-03, grad_scale: 8.0 +2023-03-21 04:58:48,056 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6718, 1.2042, 1.8752, 2.1449, 1.6733, 2.0996, 1.7747, 2.0464], + device='cuda:1'), covar=tensor([0.3247, 0.3721, 0.2188, 0.0975, 0.1926, 0.1380, 0.2255, 0.4108], + device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0072, 0.0061, 0.0053, 0.0057, 0.0056, 0.0089, 0.0058], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:58:59,896 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-21 04:59:01,629 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79026.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 04:59:05,337 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.304e+02 1.791e+02 2.019e+02 2.450e+02 4.572e+02, threshold=4.037e+02, percent-clipped=2.0 +2023-03-21 04:59:07,897 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9467, 3.5940, 3.7210, 3.7092, 3.6484, 3.4999, 3.9068, 3.5380], + device='cuda:1'), covar=tensor([0.0175, 0.0215, 0.0124, 0.0183, 0.0418, 0.0144, 0.0161, 0.0170], + device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0098, 0.0095, 0.0084, 0.0168, 0.0103, 0.0101, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 04:59:12,703 INFO [train.py:901] (1/2) Epoch 28, batch 2800, loss[loss=0.1288, simple_loss=0.2093, pruned_loss=0.02411, over 7247.00 frames. ], tot_loss[loss=0.138, simple_loss=0.2185, pruned_loss=0.02872, over 1445527.76 frames. ], batch size: 47, lr: 5.85e-03, grad_scale: 8.0 +2023-03-21 04:59:35,412 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 04:59:35,463 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 04:59:35,488 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 04:59:35,505 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 04:59:35,675 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 04:59:35,690 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 04:59:35,700 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 04:59:36,093 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 04:59:36,166 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 04:59:36,181 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 04:59:36,201 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 04:59:36,361 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 04:59:36,366 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 04:59:36,373 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 04:59:36,512 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 04:59:36,514 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 04:59:36,571 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 04:59:40,814 INFO [train.py:901] (1/2) Epoch 29, batch 0, loss[loss=0.1398, simple_loss=0.2187, pruned_loss=0.03048, over 7283.00 frames. ], tot_loss[loss=0.1398, simple_loss=0.2187, pruned_loss=0.03048, over 7283.00 frames. ], batch size: 47, lr: 5.75e-03, grad_scale: 8.0 +2023-03-21 04:59:40,815 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 05:00:02,609 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6973, 4.3229, 4.2319, 4.4850, 4.5367, 4.0599, 4.6394, 4.3799], + device='cuda:1'), covar=tensor([0.0127, 0.0150, 0.0124, 0.0123, 0.0343, 0.0144, 0.0111, 0.0125], + device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0098, 0.0095, 0.0084, 0.0168, 0.0103, 0.0101, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:00:07,206 INFO [train.py:935] (1/2) Epoch 29, validation: loss=0.1643, simple_loss=0.2541, pruned_loss=0.03725, over 1622729.00 frames. +2023-03-21 05:00:07,206 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 05:00:13,238 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 05:00:24,295 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 05:00:29,067 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79116.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:00:31,003 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 05:00:32,445 INFO [train.py:901] (1/2) Epoch 29, batch 50, loss[loss=0.1601, simple_loss=0.2352, pruned_loss=0.04253, over 7148.00 frames. ], tot_loss[loss=0.1377, simple_loss=0.2184, pruned_loss=0.02854, over 326495.26 frames. ], batch size: 98, lr: 5.75e-03, grad_scale: 16.0 +2023-03-21 05:00:33,489 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 05:00:37,126 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 05:00:38,552 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 1.920e+02 2.167e+02 2.557e+02 4.902e+02, threshold=4.333e+02, percent-clipped=3.0 +2023-03-21 05:00:53,491 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=79164.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:00:53,926 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 05:00:55,107 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 05:00:58,599 INFO [train.py:901] (1/2) Epoch 29, batch 100, loss[loss=0.1271, simple_loss=0.2115, pruned_loss=0.02136, over 7282.00 frames. ], tot_loss[loss=0.1384, simple_loss=0.2195, pruned_loss=0.02865, over 576000.00 frames. ], batch size: 68, lr: 5.75e-03, grad_scale: 16.0 +2023-03-21 05:01:18,485 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79212.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:01:20,542 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7509, 2.9430, 2.6641, 3.0068, 2.8567, 2.6411, 2.8935, 2.8169], + device='cuda:1'), covar=tensor([0.0836, 0.0813, 0.1446, 0.1531, 0.1168, 0.0633, 0.0719, 0.0994], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0054, 0.0061, 0.0053, 0.0050, 0.0055, 0.0051, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:01:24,447 INFO [train.py:901] (1/2) Epoch 29, batch 150, loss[loss=0.1342, simple_loss=0.216, pruned_loss=0.0262, over 7327.00 frames. ], tot_loss[loss=0.1393, simple_loss=0.2202, pruned_loss=0.02921, over 769811.49 frames. ], batch size: 54, lr: 5.74e-03, grad_scale: 16.0 +2023-03-21 05:01:29,961 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.271e+02 1.931e+02 2.300e+02 2.732e+02 4.086e+02, threshold=4.601e+02, percent-clipped=1.0 +2023-03-21 05:01:43,584 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=79260.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:01:44,607 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.1471, 4.6672, 4.5528, 5.0958, 4.9686, 5.0764, 4.5575, 4.6270], + device='cuda:1'), covar=tensor([0.0682, 0.2119, 0.1859, 0.0866, 0.0771, 0.1012, 0.0734, 0.1126], + device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0361, 0.0275, 0.0282, 0.0209, 0.0342, 0.0203, 0.0253], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:01:49,562 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7418, 3.0426, 3.7325, 3.8583, 4.0004, 3.9296, 3.7818, 3.7182], + device='cuda:1'), covar=tensor([0.0040, 0.0134, 0.0043, 0.0038, 0.0034, 0.0035, 0.0053, 0.0058], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0064, 0.0054, 0.0053, 0.0052, 0.0056, 0.0048, 0.0071], + device='cuda:1'), out_proj_covar=tensor([8.3567e-05, 1.3933e-04, 1.0583e-04, 9.7713e-05, 9.4832e-05, 1.0478e-04, + 1.0036e-04, 1.3940e-04], device='cuda:1') +2023-03-21 05:01:49,932 INFO [train.py:901] (1/2) Epoch 29, batch 200, loss[loss=0.1487, simple_loss=0.2332, pruned_loss=0.03208, over 7337.00 frames. ], tot_loss[loss=0.1391, simple_loss=0.22, pruned_loss=0.02906, over 921004.93 frames. ], batch size: 61, lr: 5.74e-03, grad_scale: 16.0 +2023-03-21 05:01:51,582 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2390, 4.3444, 3.5467, 3.7573, 3.3372, 2.3403, 1.9390, 4.3034], + device='cuda:1'), covar=tensor([0.0043, 0.0036, 0.0093, 0.0059, 0.0113, 0.0454, 0.0549, 0.0046], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0083, 0.0103, 0.0087, 0.0118, 0.0125, 0.0123, 0.0096], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 05:01:55,403 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 05:01:55,997 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79285.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 05:02:00,313 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 05:02:04,493 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4160, 2.2079, 2.2070, 3.5801, 1.7122, 3.3911, 1.3549, 2.8402], + device='cuda:1'), covar=tensor([0.0173, 0.1533, 0.1856, 0.0180, 0.3929, 0.0257, 0.1398, 0.0411], + device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0257, 0.0271, 0.0197, 0.0258, 0.0210, 0.0244, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:02:05,343 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 05:02:14,516 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79321.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:02:15,453 INFO [train.py:901] (1/2) Epoch 29, batch 250, loss[loss=0.1335, simple_loss=0.2225, pruned_loss=0.02219, over 7351.00 frames. ], tot_loss[loss=0.1379, simple_loss=0.2187, pruned_loss=0.02853, over 1035857.54 frames. ], batch size: 73, lr: 5.74e-03, grad_scale: 16.0 +2023-03-21 05:02:18,895 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 05:02:20,556 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79333.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:02:20,897 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.752e+02 2.021e+02 2.403e+02 4.904e+02, threshold=4.043e+02, percent-clipped=2.0 +2023-03-21 05:02:27,658 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79346.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 05:02:29,639 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2877, 3.5145, 3.4199, 3.4957, 3.2114, 3.3489, 3.6948, 3.7352], + device='cuda:1'), covar=tensor([0.0278, 0.0208, 0.0202, 0.0210, 0.0358, 0.0518, 0.0256, 0.0202], + device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0118, 0.0108, 0.0115, 0.0105, 0.0095, 0.0093, 0.0090], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:02:33,863 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.10 vs. limit=5.0 +2023-03-21 05:02:36,436 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79364.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:02:39,308 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 05:02:40,807 INFO [train.py:901] (1/2) Epoch 29, batch 300, loss[loss=0.1342, simple_loss=0.2211, pruned_loss=0.02364, over 7220.00 frames. ], tot_loss[loss=0.1383, simple_loss=0.2183, pruned_loss=0.02911, over 1121027.25 frames. ], batch size: 93, lr: 5.74e-03, grad_scale: 16.0 +2023-03-21 05:02:42,015 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79375.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 05:02:43,536 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6986, 2.3567, 2.3657, 3.8351, 1.5613, 3.4995, 1.3725, 3.0209], + device='cuda:1'), covar=tensor([0.0152, 0.1437, 0.1641, 0.0140, 0.4500, 0.0252, 0.1404, 0.0344], + device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0260, 0.0274, 0.0199, 0.0261, 0.0211, 0.0246, 0.0231], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:02:48,354 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 05:02:52,056 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79394.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:03:06,645 INFO [train.py:901] (1/2) Epoch 29, batch 350, loss[loss=0.1145, simple_loss=0.1919, pruned_loss=0.01854, over 7343.00 frames. ], tot_loss[loss=0.1391, simple_loss=0.2193, pruned_loss=0.0295, over 1191290.08 frames. ], batch size: 44, lr: 5.74e-03, grad_scale: 16.0 +2023-03-21 05:03:07,830 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79425.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:03:12,768 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.257e+02 2.121e+02 2.387e+02 2.963e+02 4.532e+02, threshold=4.773e+02, percent-clipped=2.0 +2023-03-21 05:03:13,883 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7663, 3.8265, 3.0644, 3.3712, 2.8376, 2.1284, 1.8976, 3.8273], + device='cuda:1'), covar=tensor([0.0049, 0.0046, 0.0125, 0.0071, 0.0137, 0.0510, 0.0594, 0.0060], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0083, 0.0103, 0.0087, 0.0117, 0.0124, 0.0123, 0.0095], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 05:03:13,921 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79436.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 05:03:15,818 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6068, 3.8649, 3.7270, 3.8112, 3.4397, 3.8693, 4.1886, 4.2165], + device='cuda:1'), covar=tensor([0.0274, 0.0177, 0.0207, 0.0175, 0.0410, 0.0245, 0.0193, 0.0150], + device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0118, 0.0107, 0.0114, 0.0105, 0.0094, 0.0093, 0.0090], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:03:21,278 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 05:03:32,727 INFO [train.py:901] (1/2) Epoch 29, batch 400, loss[loss=0.1224, simple_loss=0.193, pruned_loss=0.02588, over 6964.00 frames. ], tot_loss[loss=0.1389, simple_loss=0.2191, pruned_loss=0.02935, over 1245913.03 frames. ], batch size: 35, lr: 5.74e-03, grad_scale: 16.0 +2023-03-21 05:03:58,243 INFO [train.py:901] (1/2) Epoch 29, batch 450, loss[loss=0.1332, simple_loss=0.2155, pruned_loss=0.02547, over 7274.00 frames. ], tot_loss[loss=0.1392, simple_loss=0.2193, pruned_loss=0.0296, over 1291406.92 frames. ], batch size: 70, lr: 5.73e-03, grad_scale: 16.0 +2023-03-21 05:04:03,223 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 05:04:03,240 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 05:04:04,214 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.851e+02 2.031e+02 2.665e+02 3.786e+02, threshold=4.062e+02, percent-clipped=0.0 +2023-03-21 05:04:23,695 INFO [train.py:901] (1/2) Epoch 29, batch 500, loss[loss=0.138, simple_loss=0.2195, pruned_loss=0.02829, over 7356.00 frames. ], tot_loss[loss=0.1376, simple_loss=0.2177, pruned_loss=0.02877, over 1324526.55 frames. ], batch size: 51, lr: 5.73e-03, grad_scale: 8.0 +2023-03-21 05:04:29,825 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79585.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:04:36,186 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 05:04:38,036 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 05:04:38,538 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 05:04:41,608 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 05:04:45,511 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 05:04:48,602 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79621.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:04:49,484 INFO [train.py:901] (1/2) Epoch 29, batch 550, loss[loss=0.1307, simple_loss=0.2135, pruned_loss=0.02392, over 7341.00 frames. ], tot_loss[loss=0.1382, simple_loss=0.2182, pruned_loss=0.02905, over 1350520.69 frames. ], batch size: 44, lr: 5.73e-03, grad_scale: 8.0 +2023-03-21 05:04:55,504 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.766e+02 2.094e+02 2.391e+02 9.522e+02, threshold=4.188e+02, percent-clipped=4.0 +2023-03-21 05:04:55,549 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 05:04:58,575 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79641.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 05:05:01,137 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79646.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:05:04,521 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 05:05:07,042 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79657.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:05:07,454 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 05:05:13,019 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=79669.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:05:14,473 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 05:05:14,957 INFO [train.py:901] (1/2) Epoch 29, batch 600, loss[loss=0.1508, simple_loss=0.2298, pruned_loss=0.03595, over 7302.00 frames. ], tot_loss[loss=0.139, simple_loss=0.2189, pruned_loss=0.02961, over 1370228.29 frames. ], batch size: 86, lr: 5.73e-03, grad_scale: 8.0 +2023-03-21 05:05:23,041 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79689.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:05:30,137 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7482, 4.1199, 4.4148, 4.4790, 4.3868, 4.3226, 4.7473, 4.0930], + device='cuda:1'), covar=tensor([0.0111, 0.0154, 0.0105, 0.0103, 0.0307, 0.0092, 0.0091, 0.0125], + device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0095, 0.0091, 0.0081, 0.0161, 0.0100, 0.0097, 0.0101], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:05:31,992 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 05:05:38,051 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79718.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:05:39,008 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79720.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:05:40,448 INFO [train.py:901] (1/2) Epoch 29, batch 650, loss[loss=0.1374, simple_loss=0.2203, pruned_loss=0.02729, over 7328.00 frames. ], tot_loss[loss=0.1387, simple_loss=0.2188, pruned_loss=0.02929, over 1385897.76 frames. ], batch size: 61, lr: 5.73e-03, grad_scale: 8.0 +2023-03-21 05:05:40,947 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 05:05:41,653 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-03-21 05:05:44,413 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79731.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 05:05:46,324 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.341e+02 1.881e+02 2.206e+02 2.571e+02 4.157e+02, threshold=4.413e+02, percent-clipped=0.0 +2023-03-21 05:05:58,454 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 05:06:00,588 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79762.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:06:05,893 INFO [train.py:901] (1/2) Epoch 29, batch 700, loss[loss=0.1502, simple_loss=0.2338, pruned_loss=0.0333, over 7290.00 frames. ], tot_loss[loss=0.139, simple_loss=0.219, pruned_loss=0.02947, over 1399869.39 frames. ], batch size: 86, lr: 5.72e-03, grad_scale: 8.0 +2023-03-21 05:06:06,411 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 05:06:23,314 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 +2023-03-21 05:06:31,466 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 05:06:31,977 INFO [train.py:901] (1/2) Epoch 29, batch 750, loss[loss=0.1419, simple_loss=0.2206, pruned_loss=0.03164, over 7372.00 frames. ], tot_loss[loss=0.1389, simple_loss=0.2189, pruned_loss=0.0294, over 1409765.88 frames. ], batch size: 65, lr: 5.72e-03, grad_scale: 8.0 +2023-03-21 05:06:31,984 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 05:06:32,113 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79823.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:06:33,761 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8411, 2.1374, 1.7280, 2.6700, 2.6659, 2.6470, 2.2054, 2.3977], + device='cuda:1'), covar=tensor([0.1830, 0.0952, 0.3325, 0.0796, 0.0264, 0.0265, 0.0323, 0.0378], + device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0231, 0.0255, 0.0261, 0.0187, 0.0182, 0.0207, 0.0219], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 05:06:38,568 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.855e+02 2.149e+02 2.429e+02 6.154e+02, threshold=4.298e+02, percent-clipped=2.0 +2023-03-21 05:06:40,948 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 +2023-03-21 05:06:46,674 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 05:06:52,080 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 05:06:57,599 INFO [train.py:901] (1/2) Epoch 29, batch 800, loss[loss=0.149, simple_loss=0.2289, pruned_loss=0.03454, over 7312.00 frames. ], tot_loss[loss=0.1387, simple_loss=0.2188, pruned_loss=0.02934, over 1418306.86 frames. ], batch size: 59, lr: 5.72e-03, grad_scale: 8.0 +2023-03-21 05:06:58,724 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 05:06:59,740 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 05:07:10,169 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 05:07:10,515 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-03-21 05:07:15,239 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9619, 3.4377, 3.9123, 4.0725, 4.0943, 4.0547, 4.1175, 3.9714], + device='cuda:1'), covar=tensor([0.0034, 0.0106, 0.0034, 0.0029, 0.0030, 0.0033, 0.0031, 0.0046], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0063, 0.0053, 0.0052, 0.0051, 0.0056, 0.0047, 0.0070], + device='cuda:1'), out_proj_covar=tensor([8.2336e-05, 1.3705e-04, 1.0445e-04, 9.5530e-05, 9.3374e-05, 1.0327e-04, + 9.8010e-05, 1.3673e-04], device='cuda:1') +2023-03-21 05:07:15,801 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5563, 3.0428, 3.3627, 3.4435, 2.8814, 2.7137, 3.4545, 2.5476], + device='cuda:1'), covar=tensor([0.0438, 0.0567, 0.0605, 0.0574, 0.0609, 0.0815, 0.0620, 0.1615], + device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0339, 0.0273, 0.0357, 0.0298, 0.0296, 0.0347, 0.0267], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 05:07:23,766 INFO [train.py:901] (1/2) Epoch 29, batch 850, loss[loss=0.144, simple_loss=0.2204, pruned_loss=0.03376, over 7293.00 frames. ], tot_loss[loss=0.1395, simple_loss=0.2193, pruned_loss=0.02981, over 1424265.34 frames. ], batch size: 77, lr: 5.72e-03, grad_scale: 8.0 +2023-03-21 05:07:29,315 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 05:07:29,324 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 05:07:29,786 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.291e+02 1.737e+02 2.030e+02 2.344e+02 3.800e+02, threshold=4.059e+02, percent-clipped=0.0 +2023-03-21 05:07:32,890 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79941.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:07:32,916 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79941.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 05:07:34,859 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 05:07:37,109 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-03-21 05:07:38,426 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 05:07:49,413 INFO [train.py:901] (1/2) Epoch 29, batch 900, loss[loss=0.1297, simple_loss=0.2129, pruned_loss=0.02321, over 7301.00 frames. ], tot_loss[loss=0.1388, simple_loss=0.2188, pruned_loss=0.02943, over 1429771.48 frames. ], batch size: 49, lr: 5.72e-03, grad_scale: 8.0 +2023-03-21 05:07:57,538 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=79989.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 05:07:57,566 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79989.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:08:13,984 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80013.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:08:17,421 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80020.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:08:18,837 INFO [train.py:901] (1/2) Epoch 29, batch 950, loss[loss=0.1415, simple_loss=0.2167, pruned_loss=0.03319, over 7259.00 frames. ], tot_loss[loss=0.1379, simple_loss=0.218, pruned_loss=0.02889, over 1431755.68 frames. ], batch size: 47, lr: 5.72e-03, grad_scale: 8.0 +2023-03-21 05:08:20,840 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 05:08:22,969 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80031.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 05:08:24,817 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.204e+02 1.694e+02 2.042e+02 2.334e+02 4.192e+02, threshold=4.084e+02, percent-clipped=1.0 +2023-03-21 05:08:25,855 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=80037.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:08:41,921 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=80068.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:08:43,918 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 05:08:44,391 INFO [train.py:901] (1/2) Epoch 29, batch 1000, loss[loss=0.1289, simple_loss=0.2106, pruned_loss=0.02359, over 7301.00 frames. ], tot_loss[loss=0.1386, simple_loss=0.2188, pruned_loss=0.02919, over 1433678.31 frames. ], batch size: 68, lr: 5.71e-03, grad_scale: 8.0 +2023-03-21 05:08:47,417 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=80079.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 05:08:53,238 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 +2023-03-21 05:09:02,080 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8751, 3.0984, 2.6073, 3.0422, 2.9189, 2.7293, 2.9618, 2.8189], + device='cuda:1'), covar=tensor([0.0599, 0.0522, 0.0993, 0.0581, 0.1120, 0.0451, 0.0669, 0.0781], + device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0054, 0.0061, 0.0053, 0.0052, 0.0055, 0.0052, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:09:05,379 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 05:09:07,465 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80118.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:09:09,826 INFO [train.py:901] (1/2) Epoch 29, batch 1050, loss[loss=0.1568, simple_loss=0.2316, pruned_loss=0.04102, over 7303.00 frames. ], tot_loss[loss=0.1392, simple_loss=0.2193, pruned_loss=0.02951, over 1435910.08 frames. ], batch size: 49, lr: 5.71e-03, grad_scale: 8.0 +2023-03-21 05:09:16,379 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.326e+02 1.871e+02 2.139e+02 2.585e+02 3.936e+02, threshold=4.278e+02, percent-clipped=0.0 +2023-03-21 05:09:26,991 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 05:09:30,989 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 05:09:35,404 INFO [train.py:901] (1/2) Epoch 29, batch 1100, loss[loss=0.1421, simple_loss=0.2193, pruned_loss=0.03251, over 7358.00 frames. ], tot_loss[loss=0.1383, simple_loss=0.2185, pruned_loss=0.02906, over 1437346.14 frames. ], batch size: 73, lr: 5.71e-03, grad_scale: 8.0 +2023-03-21 05:09:54,053 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-03-21 05:09:59,453 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 05:09:59,972 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 05:10:01,980 INFO [train.py:901] (1/2) Epoch 29, batch 1150, loss[loss=0.1411, simple_loss=0.2197, pruned_loss=0.03127, over 7206.00 frames. ], tot_loss[loss=0.1384, simple_loss=0.2186, pruned_loss=0.02913, over 1436691.14 frames. ], batch size: 50, lr: 5.71e-03, grad_scale: 8.0 +2023-03-21 05:10:05,380 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-03-21 05:10:08,130 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.333e+02 1.785e+02 2.047e+02 2.431e+02 3.454e+02, threshold=4.094e+02, percent-clipped=0.0 +2023-03-21 05:10:10,691 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 05:10:11,228 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 05:10:11,300 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80241.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:10:24,732 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0055, 4.2063, 3.9889, 4.1826, 3.8437, 4.1981, 4.4965, 4.5715], + device='cuda:1'), covar=tensor([0.0210, 0.0132, 0.0183, 0.0133, 0.0310, 0.0225, 0.0195, 0.0136], + device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0118, 0.0108, 0.0115, 0.0106, 0.0095, 0.0092, 0.0091], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:10:27,688 INFO [train.py:901] (1/2) Epoch 29, batch 1200, loss[loss=0.1262, simple_loss=0.2149, pruned_loss=0.01875, over 7276.00 frames. ], tot_loss[loss=0.1388, simple_loss=0.2189, pruned_loss=0.02939, over 1437732.40 frames. ], batch size: 70, lr: 5.71e-03, grad_scale: 8.0 +2023-03-21 05:10:32,839 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80283.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:10:35,732 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=80289.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:10:40,918 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8775, 3.2784, 3.7435, 3.9060, 3.9334, 3.8925, 3.9182, 3.8740], + device='cuda:1'), covar=tensor([0.0026, 0.0102, 0.0035, 0.0027, 0.0026, 0.0028, 0.0042, 0.0039], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0064, 0.0055, 0.0053, 0.0052, 0.0056, 0.0048, 0.0071], + device='cuda:1'), out_proj_covar=tensor([8.1910e-05, 1.3840e-04, 1.0726e-04, 9.7731e-05, 9.3927e-05, 1.0480e-04, + 1.0013e-04, 1.3887e-04], device='cuda:1') +2023-03-21 05:10:43,313 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 05:10:45,436 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2045, 2.0780, 2.4179, 2.2849, 2.6610, 2.1914, 2.1351, 1.9811], + device='cuda:1'), covar=tensor([0.0418, 0.0661, 0.0216, 0.0196, 0.0377, 0.0803, 0.0438, 0.0300], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0033, 0.0034, 0.0031, 0.0031, 0.0031, 0.0035, 0.0035], + device='cuda:1'), out_proj_covar=tensor([8.6896e-05, 8.5679e-05, 8.3605e-05, 8.0503e-05, 8.2317e-05, 8.0756e-05, + 8.6737e-05, 8.9487e-05], device='cuda:1') +2023-03-21 05:10:48,354 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80313.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:10:53,178 INFO [train.py:901] (1/2) Epoch 29, batch 1250, loss[loss=0.1435, simple_loss=0.2228, pruned_loss=0.03204, over 7214.00 frames. ], tot_loss[loss=0.1389, simple_loss=0.2192, pruned_loss=0.02927, over 1440086.55 frames. ], batch size: 50, lr: 5.70e-03, grad_scale: 8.0 +2023-03-21 05:10:59,177 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.228e+02 1.776e+02 2.060e+02 2.370e+02 4.071e+02, threshold=4.121e+02, percent-clipped=0.0 +2023-03-21 05:11:00,299 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.9064, 4.4140, 4.2559, 4.8674, 4.6931, 4.8363, 4.2787, 4.4305], + device='cuda:1'), covar=tensor([0.0847, 0.2732, 0.2442, 0.1039, 0.1001, 0.1125, 0.0832, 0.1090], + device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0372, 0.0283, 0.0294, 0.0217, 0.0351, 0.0211, 0.0261], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:11:03,945 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80344.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:11:06,357 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 05:11:10,409 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 05:11:11,902 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 05:11:12,927 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=80361.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:11:19,042 INFO [train.py:901] (1/2) Epoch 29, batch 1300, loss[loss=0.143, simple_loss=0.2253, pruned_loss=0.03038, over 7360.00 frames. ], tot_loss[loss=0.1384, simple_loss=0.2188, pruned_loss=0.02902, over 1440214.93 frames. ], batch size: 54, lr: 5.70e-03, grad_scale: 8.0 +2023-03-21 05:11:29,865 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80393.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:11:36,164 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 05:11:38,162 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 05:11:41,588 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 05:11:42,734 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80418.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:11:45,132 INFO [train.py:901] (1/2) Epoch 29, batch 1350, loss[loss=0.1758, simple_loss=0.2517, pruned_loss=0.04995, over 6588.00 frames. ], tot_loss[loss=0.1385, simple_loss=0.219, pruned_loss=0.029, over 1440745.06 frames. ], batch size: 106, lr: 5.70e-03, grad_scale: 8.0 +2023-03-21 05:11:48,301 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9955, 3.0340, 2.4405, 3.9265, 1.8827, 3.7272, 1.5665, 3.2248], + device='cuda:1'), covar=tensor([0.0135, 0.0962, 0.1649, 0.0120, 0.3856, 0.0186, 0.1185, 0.0351], + device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0257, 0.0270, 0.0198, 0.0255, 0.0210, 0.0242, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:11:51,636 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.192e+02 1.765e+02 2.078e+02 2.410e+02 4.380e+02, threshold=4.156e+02, percent-clipped=1.0 +2023-03-21 05:11:52,161 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 05:12:01,239 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80454.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:12:05,648 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80463.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:12:07,007 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=80466.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:12:10,974 INFO [train.py:901] (1/2) Epoch 29, batch 1400, loss[loss=0.1415, simple_loss=0.2157, pruned_loss=0.03367, over 7282.00 frames. ], tot_loss[loss=0.1378, simple_loss=0.2184, pruned_loss=0.02857, over 1441363.43 frames. ], batch size: 47, lr: 5.70e-03, grad_scale: 8.0 +2023-03-21 05:12:21,788 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80494.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:12:24,679 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 05:12:32,687 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0898, 3.2506, 4.0707, 4.0956, 4.1797, 4.0568, 4.2411, 4.1057], + device='cuda:1'), covar=tensor([0.0030, 0.0124, 0.0032, 0.0029, 0.0029, 0.0032, 0.0027, 0.0044], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0063, 0.0054, 0.0053, 0.0051, 0.0056, 0.0047, 0.0070], + device='cuda:1'), out_proj_covar=tensor([8.1580e-05, 1.3675e-04, 1.0519e-04, 9.7877e-05, 9.3372e-05, 1.0372e-04, + 9.8781e-05, 1.3705e-04], device='cuda:1') +2023-03-21 05:12:36,600 INFO [train.py:901] (1/2) Epoch 29, batch 1450, loss[loss=0.1303, simple_loss=0.2072, pruned_loss=0.02669, over 7261.00 frames. ], tot_loss[loss=0.1379, simple_loss=0.2184, pruned_loss=0.02869, over 1441167.32 frames. ], batch size: 64, lr: 5.70e-03, grad_scale: 8.0 +2023-03-21 05:12:37,257 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80524.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:12:37,471 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 +2023-03-21 05:12:42,703 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.303e+02 1.883e+02 2.179e+02 2.600e+02 4.285e+02, threshold=4.357e+02, percent-clipped=2.0 +2023-03-21 05:12:48,758 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 05:12:53,542 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80555.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:12:55,958 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3896, 2.6957, 3.3419, 3.4307, 3.4965, 3.5230, 3.2995, 3.3503], + device='cuda:1'), covar=tensor([0.0036, 0.0149, 0.0046, 0.0041, 0.0036, 0.0034, 0.0094, 0.0058], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0062, 0.0053, 0.0052, 0.0050, 0.0055, 0.0047, 0.0069], + device='cuda:1'), out_proj_covar=tensor([8.0548e-05, 1.3484e-04, 1.0372e-04, 9.6137e-05, 9.1842e-05, 1.0203e-04, + 9.7250e-05, 1.3494e-04], device='cuda:1') +2023-03-21 05:13:02,255 INFO [train.py:901] (1/2) Epoch 29, batch 1500, loss[loss=0.1426, simple_loss=0.2239, pruned_loss=0.03069, over 7285.00 frames. ], tot_loss[loss=0.1379, simple_loss=0.2186, pruned_loss=0.02856, over 1440838.84 frames. ], batch size: 66, lr: 5.70e-03, grad_scale: 8.0 +2023-03-21 05:13:04,810 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 05:13:28,296 INFO [train.py:901] (1/2) Epoch 29, batch 1550, loss[loss=0.1424, simple_loss=0.2308, pruned_loss=0.02704, over 7294.00 frames. ], tot_loss[loss=0.1377, simple_loss=0.2184, pruned_loss=0.0285, over 1441001.37 frames. ], batch size: 80, lr: 5.69e-03, grad_scale: 8.0 +2023-03-21 05:13:29,865 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 05:13:34,383 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.279e+02 1.859e+02 2.000e+02 2.568e+02 5.498e+02, threshold=3.999e+02, percent-clipped=3.0 +2023-03-21 05:13:35,710 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 +2023-03-21 05:13:36,483 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80639.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:13:42,830 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-03-21 05:13:54,015 INFO [train.py:901] (1/2) Epoch 29, batch 1600, loss[loss=0.1359, simple_loss=0.2234, pruned_loss=0.02423, over 7334.00 frames. ], tot_loss[loss=0.1368, simple_loss=0.2175, pruned_loss=0.02805, over 1441686.24 frames. ], batch size: 75, lr: 5.69e-03, grad_scale: 8.0 +2023-03-21 05:14:01,549 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 05:14:02,059 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 05:14:02,162 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6179, 4.1345, 4.1753, 4.2432, 4.1927, 4.1041, 4.4592, 3.9899], + device='cuda:1'), covar=tensor([0.0139, 0.0159, 0.0130, 0.0155, 0.0384, 0.0129, 0.0154, 0.0146], + device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0095, 0.0092, 0.0081, 0.0163, 0.0100, 0.0098, 0.0101], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:14:05,026 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 05:14:15,107 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 05:14:19,249 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 05:14:19,734 INFO [train.py:901] (1/2) Epoch 29, batch 1650, loss[loss=0.1608, simple_loss=0.2411, pruned_loss=0.04024, over 7272.00 frames. ], tot_loss[loss=0.1363, simple_loss=0.2169, pruned_loss=0.02784, over 1440350.36 frames. ], batch size: 66, lr: 5.69e-03, grad_scale: 8.0 +2023-03-21 05:14:26,390 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.202e+02 1.728e+02 1.955e+02 2.318e+02 5.387e+02, threshold=3.909e+02, percent-clipped=3.0 +2023-03-21 05:14:27,082 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80736.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:14:27,997 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 05:14:33,607 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80749.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:14:33,633 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4491, 3.5791, 3.3261, 3.5333, 3.3480, 3.4446, 3.8679, 3.8941], + device='cuda:1'), covar=tensor([0.0281, 0.0206, 0.0306, 0.0221, 0.0397, 0.0516, 0.0244, 0.0188], + device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0118, 0.0109, 0.0116, 0.0106, 0.0096, 0.0092, 0.0091], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:14:44,936 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 05:14:45,409 INFO [train.py:901] (1/2) Epoch 29, batch 1700, loss[loss=0.1288, simple_loss=0.2137, pruned_loss=0.02196, over 7283.00 frames. ], tot_loss[loss=0.1374, simple_loss=0.2182, pruned_loss=0.02828, over 1442822.61 frames. ], batch size: 70, lr: 5.69e-03, grad_scale: 8.0 +2023-03-21 05:14:45,513 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80773.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:14:49,994 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 05:14:58,100 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80797.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:15:00,872 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 05:15:09,976 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80819.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:15:11,910 INFO [train.py:901] (1/2) Epoch 29, batch 1750, loss[loss=0.1381, simple_loss=0.2213, pruned_loss=0.02742, over 7352.00 frames. ], tot_loss[loss=0.1376, simple_loss=0.218, pruned_loss=0.02862, over 1442415.71 frames. ], batch size: 63, lr: 5.69e-03, grad_scale: 8.0 +2023-03-21 05:15:17,527 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80834.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:15:17,862 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.159e+02 1.786e+02 2.210e+02 2.520e+02 4.800e+02, threshold=4.419e+02, percent-clipped=3.0 +2023-03-21 05:15:24,284 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 05:15:25,292 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 05:15:25,338 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80850.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:15:37,356 INFO [train.py:901] (1/2) Epoch 29, batch 1800, loss[loss=0.1353, simple_loss=0.2116, pruned_loss=0.02949, over 7266.00 frames. ], tot_loss[loss=0.1379, simple_loss=0.2184, pruned_loss=0.02874, over 1445155.03 frames. ], batch size: 52, lr: 5.69e-03, grad_scale: 8.0 +2023-03-21 05:15:42,986 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2920, 3.3029, 3.5287, 2.9856, 3.2694, 3.2936, 3.0278, 3.4019], + device='cuda:1'), covar=tensor([0.1637, 0.0605, 0.1098, 0.1755, 0.0919, 0.1035, 0.1625, 0.1244], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0062, 0.0048, 0.0047, 0.0046, 0.0045, 0.0064, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:15:44,559 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4206, 3.3687, 2.3829, 3.8982, 2.8174, 3.4299, 1.5404, 2.2242], + device='cuda:1'), covar=tensor([0.0410, 0.0786, 0.2829, 0.0499, 0.0439, 0.0518, 0.3846, 0.2335], + device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0260, 0.0292, 0.0272, 0.0274, 0.0272, 0.0250, 0.0273], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:15:47,392 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 05:16:00,776 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 05:16:02,654 INFO [train.py:901] (1/2) Epoch 29, batch 1850, loss[loss=0.1511, simple_loss=0.2247, pruned_loss=0.0387, over 7271.00 frames. ], tot_loss[loss=0.1379, simple_loss=0.2182, pruned_loss=0.02877, over 1444573.62 frames. ], batch size: 66, lr: 5.68e-03, grad_scale: 8.0 +2023-03-21 05:16:08,667 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.303e+02 1.763e+02 2.062e+02 2.438e+02 5.216e+02, threshold=4.123e+02, percent-clipped=1.0 +2023-03-21 05:16:10,681 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 05:16:10,800 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80939.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:16:13,904 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2949, 3.4413, 3.3010, 3.4426, 3.2843, 3.2791, 3.7116, 3.7216], + device='cuda:1'), covar=tensor([0.0263, 0.0206, 0.0242, 0.0211, 0.0306, 0.0601, 0.0217, 0.0191], + device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0118, 0.0108, 0.0114, 0.0105, 0.0094, 0.0091, 0.0091], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:16:20,463 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9081, 2.9439, 2.6839, 2.8238, 2.8324, 2.7362, 3.0143, 2.7939], + device='cuda:1'), covar=tensor([0.0480, 0.1313, 0.0888, 0.1497, 0.1866, 0.0751, 0.0711, 0.1466], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0053, 0.0061, 0.0053, 0.0051, 0.0055, 0.0052, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:16:23,961 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2390, 4.2892, 3.7550, 3.7885, 3.4529, 2.5075, 1.9624, 4.4384], + device='cuda:1'), covar=tensor([0.0089, 0.0060, 0.0132, 0.0084, 0.0163, 0.0589, 0.0667, 0.0071], + device='cuda:1'), in_proj_covar=tensor([0.0104, 0.0086, 0.0105, 0.0090, 0.0120, 0.0126, 0.0125, 0.0097], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 05:16:27,919 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 05:16:28,401 INFO [train.py:901] (1/2) Epoch 29, batch 1900, loss[loss=0.1277, simple_loss=0.2137, pruned_loss=0.02082, over 7361.00 frames. ], tot_loss[loss=0.1372, simple_loss=0.2175, pruned_loss=0.02849, over 1443177.87 frames. ], batch size: 73, lr: 5.68e-03, grad_scale: 8.0 +2023-03-21 05:16:35,368 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=80987.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:16:52,852 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 05:16:54,333 INFO [train.py:901] (1/2) Epoch 29, batch 1950, loss[loss=0.1406, simple_loss=0.2188, pruned_loss=0.03124, over 7288.00 frames. ], tot_loss[loss=0.1373, simple_loss=0.2176, pruned_loss=0.02852, over 1442995.13 frames. ], batch size: 68, lr: 5.68e-03, grad_scale: 8.0 +2023-03-21 05:17:00,348 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.014e+02 1.770e+02 2.041e+02 2.360e+02 4.145e+02, threshold=4.083e+02, percent-clipped=1.0 +2023-03-21 05:17:03,411 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 05:17:08,042 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81049.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:17:08,431 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 05:17:08,960 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 05:17:19,948 INFO [train.py:901] (1/2) Epoch 29, batch 2000, loss[loss=0.1512, simple_loss=0.2304, pruned_loss=0.03599, over 7255.00 frames. ], tot_loss[loss=0.1371, simple_loss=0.2177, pruned_loss=0.02822, over 1445712.11 frames. ], batch size: 64, lr: 5.68e-03, grad_scale: 8.0 +2023-03-21 05:17:25,562 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 05:17:30,200 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81092.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:17:32,717 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=81097.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:17:36,638 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 05:17:42,280 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81116.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:17:43,785 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81119.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:17:44,193 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 05:17:45,680 INFO [train.py:901] (1/2) Epoch 29, batch 2050, loss[loss=0.1344, simple_loss=0.2172, pruned_loss=0.02581, over 7360.00 frames. ], tot_loss[loss=0.1365, simple_loss=0.2171, pruned_loss=0.02799, over 1445733.14 frames. ], batch size: 65, lr: 5.68e-03, grad_scale: 8.0 +2023-03-21 05:17:48,728 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81129.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:17:52,345 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.043e+02 1.747e+02 2.128e+02 2.530e+02 3.662e+02, threshold=4.256e+02, percent-clipped=0.0 +2023-03-21 05:18:00,128 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81150.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:18:08,564 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=81167.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:18:12,083 INFO [train.py:901] (1/2) Epoch 29, batch 2100, loss[loss=0.1325, simple_loss=0.2118, pruned_loss=0.02654, over 7299.00 frames. ], tot_loss[loss=0.1363, simple_loss=0.2168, pruned_loss=0.02792, over 1444696.84 frames. ], batch size: 70, lr: 5.67e-03, grad_scale: 8.0 +2023-03-21 05:18:14,171 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81177.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:18:19,076 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 05:18:22,079 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 05:18:24,653 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=81198.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:18:25,217 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9484, 2.7003, 3.2605, 2.9072, 3.2305, 2.9000, 2.5700, 2.9795], + device='cuda:1'), covar=tensor([0.2046, 0.1018, 0.1344, 0.1403, 0.1092, 0.1191, 0.2445, 0.1555], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0063, 0.0048, 0.0048, 0.0047, 0.0045, 0.0065, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 05:18:31,205 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.26 vs. limit=5.0 +2023-03-21 05:18:36,534 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4028, 4.0273, 4.4800, 4.3787, 4.4415, 4.4906, 4.4945, 4.1433], + device='cuda:1'), covar=tensor([0.0037, 0.0112, 0.0036, 0.0039, 0.0036, 0.0038, 0.0032, 0.0063], + device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0063, 0.0054, 0.0052, 0.0051, 0.0056, 0.0047, 0.0070], + device='cuda:1'), out_proj_covar=tensor([8.3380e-05, 1.3737e-04, 1.0548e-04, 9.6177e-05, 9.3823e-05, 1.0270e-04, + 9.8177e-05, 1.3553e-04], device='cuda:1') +2023-03-21 05:18:37,951 INFO [train.py:901] (1/2) Epoch 29, batch 2150, loss[loss=0.1368, simple_loss=0.2163, pruned_loss=0.02862, over 7330.00 frames. ], tot_loss[loss=0.1368, simple_loss=0.2174, pruned_loss=0.02808, over 1444165.35 frames. ], batch size: 49, lr: 5.67e-03, grad_scale: 8.0 +2023-03-21 05:18:43,969 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.793e+02 2.082e+02 2.494e+02 7.569e+02, threshold=4.163e+02, percent-clipped=1.0 +2023-03-21 05:18:46,425 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 +2023-03-21 05:18:57,461 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0224, 3.6270, 3.7265, 3.7285, 3.5956, 3.5957, 3.8897, 3.4245], + device='cuda:1'), covar=tensor([0.0134, 0.0204, 0.0118, 0.0164, 0.0444, 0.0111, 0.0171, 0.0196], + device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0096, 0.0093, 0.0083, 0.0165, 0.0101, 0.0100, 0.0102], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:19:03,834 INFO [train.py:901] (1/2) Epoch 29, batch 2200, loss[loss=0.1505, simple_loss=0.2291, pruned_loss=0.03597, over 7321.00 frames. ], tot_loss[loss=0.1361, simple_loss=0.2166, pruned_loss=0.02783, over 1444010.11 frames. ], batch size: 61, lr: 5.67e-03, grad_scale: 8.0 +2023-03-21 05:19:09,306 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 05:19:16,924 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81299.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:19:17,909 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81301.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:19:29,335 INFO [train.py:901] (1/2) Epoch 29, batch 2250, loss[loss=0.1601, simple_loss=0.2349, pruned_loss=0.04267, over 7251.00 frames. ], tot_loss[loss=0.1368, simple_loss=0.2171, pruned_loss=0.02826, over 1443358.04 frames. ], batch size: 55, lr: 5.67e-03, grad_scale: 8.0 +2023-03-21 05:19:35,476 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 1.800e+02 2.074e+02 2.534e+02 4.493e+02, threshold=4.147e+02, percent-clipped=2.0 +2023-03-21 05:19:44,689 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 05:19:45,186 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 05:19:48,862 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81360.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:19:49,831 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81362.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:19:55,266 INFO [train.py:901] (1/2) Epoch 29, batch 2300, loss[loss=0.1484, simple_loss=0.2249, pruned_loss=0.03591, over 7272.00 frames. ], tot_loss[loss=0.1372, simple_loss=0.2174, pruned_loss=0.02845, over 1445066.97 frames. ], batch size: 66, lr: 5.67e-03, grad_scale: 8.0 +2023-03-21 05:19:57,256 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 05:20:04,851 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81392.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:20:12,946 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6033, 4.1065, 3.9703, 4.5463, 4.4001, 4.4851, 3.7961, 4.0932], + device='cuda:1'), covar=tensor([0.0773, 0.2604, 0.2480, 0.1209, 0.0882, 0.1211, 0.0946, 0.1106], + device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0374, 0.0283, 0.0293, 0.0215, 0.0352, 0.0214, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:20:21,608 INFO [train.py:901] (1/2) Epoch 29, batch 2350, loss[loss=0.1602, simple_loss=0.243, pruned_loss=0.03873, over 6780.00 frames. ], tot_loss[loss=0.1374, simple_loss=0.2176, pruned_loss=0.02857, over 1443332.14 frames. ], batch size: 107, lr: 5.67e-03, grad_scale: 8.0 +2023-03-21 05:20:24,696 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81429.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:20:26,272 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8069, 3.4383, 3.3212, 3.6168, 3.1715, 2.8346, 3.7031, 2.7464], + device='cuda:1'), covar=tensor([0.0435, 0.0483, 0.0612, 0.0530, 0.0919, 0.1196, 0.0563, 0.1999], + device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0340, 0.0271, 0.0355, 0.0299, 0.0298, 0.0346, 0.0268], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 05:20:27,536 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.392e+02 1.900e+02 2.109e+02 2.477e+02 4.318e+02, threshold=4.218e+02, percent-clipped=1.0 +2023-03-21 05:20:30,784 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=81440.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:20:34,444 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2683, 2.6364, 1.9886, 3.2392, 2.9581, 3.2864, 2.0987, 2.8484], + device='cuda:1'), covar=tensor([0.1922, 0.0963, 0.3316, 0.0450, 0.0206, 0.0209, 0.0299, 0.0315], + device='cuda:1'), in_proj_covar=tensor([0.0259, 0.0231, 0.0259, 0.0265, 0.0189, 0.0185, 0.0210, 0.0222], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 05:20:44,573 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1293, 3.0763, 2.2889, 3.6323, 2.3379, 3.2244, 1.4328, 2.0941], + device='cuda:1'), covar=tensor([0.0393, 0.0762, 0.2390, 0.0589, 0.0432, 0.0512, 0.3569, 0.1627], + device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0259, 0.0289, 0.0271, 0.0273, 0.0269, 0.0248, 0.0269], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:20:44,922 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 05:20:45,272 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 +2023-03-21 05:20:46,987 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81472.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:20:47,396 INFO [train.py:901] (1/2) Epoch 29, batch 2400, loss[loss=0.1355, simple_loss=0.2159, pruned_loss=0.02754, over 7318.00 frames. ], tot_loss[loss=0.1372, simple_loss=0.2177, pruned_loss=0.02835, over 1444238.48 frames. ], batch size: 49, lr: 5.66e-03, grad_scale: 8.0 +2023-03-21 05:20:49,480 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=81477.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:20:51,391 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 05:21:02,490 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 05:21:05,005 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 05:21:13,055 INFO [train.py:901] (1/2) Epoch 29, batch 2450, loss[loss=0.1374, simple_loss=0.218, pruned_loss=0.0284, over 7314.00 frames. ], tot_loss[loss=0.1372, simple_loss=0.2179, pruned_loss=0.02828, over 1444767.23 frames. ], batch size: 80, lr: 5.66e-03, grad_scale: 8.0 +2023-03-21 05:21:19,660 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.288e+02 1.787e+02 2.100e+02 2.538e+02 7.882e+02, threshold=4.199e+02, percent-clipped=1.0 +2023-03-21 05:21:31,140 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 05:21:39,151 INFO [train.py:901] (1/2) Epoch 29, batch 2500, loss[loss=0.1307, simple_loss=0.2056, pruned_loss=0.02787, over 7124.00 frames. ], tot_loss[loss=0.1383, simple_loss=0.2188, pruned_loss=0.02885, over 1444624.43 frames. ], batch size: 41, lr: 5.66e-03, grad_scale: 16.0 +2023-03-21 05:21:57,435 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 05:22:05,110 INFO [train.py:901] (1/2) Epoch 29, batch 2550, loss[loss=0.1606, simple_loss=0.2422, pruned_loss=0.03951, over 6696.00 frames. ], tot_loss[loss=0.1379, simple_loss=0.2183, pruned_loss=0.02878, over 1443167.71 frames. ], batch size: 106, lr: 5.66e-03, grad_scale: 16.0 +2023-03-21 05:22:11,198 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.193e+02 1.737e+02 2.185e+02 2.560e+02 4.324e+02, threshold=4.370e+02, percent-clipped=1.0 +2023-03-21 05:22:21,895 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81655.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:22:22,873 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81657.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:22:29,457 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81670.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:22:30,798 INFO [train.py:901] (1/2) Epoch 29, batch 2600, loss[loss=0.1308, simple_loss=0.215, pruned_loss=0.02323, over 7297.00 frames. ], tot_loss[loss=0.138, simple_loss=0.2185, pruned_loss=0.02872, over 1444309.89 frames. ], batch size: 68, lr: 5.66e-03, grad_scale: 16.0 +2023-03-21 05:22:55,364 INFO [train.py:901] (1/2) Epoch 29, batch 2650, loss[loss=0.1445, simple_loss=0.2196, pruned_loss=0.03474, over 7305.00 frames. ], tot_loss[loss=0.1384, simple_loss=0.2187, pruned_loss=0.02907, over 1443112.85 frames. ], batch size: 49, lr: 5.66e-03, grad_scale: 8.0 +2023-03-21 05:22:59,393 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81731.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 05:23:01,693 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.150e+02 1.754e+02 2.071e+02 2.481e+02 5.009e+02, threshold=4.143e+02, percent-clipped=1.0 +2023-03-21 05:23:20,013 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81772.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:23:20,403 INFO [train.py:901] (1/2) Epoch 29, batch 2700, loss[loss=0.1282, simple_loss=0.215, pruned_loss=0.02068, over 7340.00 frames. ], tot_loss[loss=0.1372, simple_loss=0.2176, pruned_loss=0.02843, over 1442740.79 frames. ], batch size: 73, lr: 5.65e-03, grad_scale: 8.0 +2023-03-21 05:23:37,493 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5073, 5.0169, 5.0810, 5.0221, 4.8890, 4.5649, 5.1076, 4.9524], + device='cuda:1'), covar=tensor([0.0430, 0.0362, 0.0339, 0.0446, 0.0299, 0.0317, 0.0310, 0.0385], + device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0238, 0.0181, 0.0187, 0.0148, 0.0217, 0.0194, 0.0142], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 05:23:43,949 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=81820.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:23:45,381 INFO [train.py:901] (1/2) Epoch 29, batch 2750, loss[loss=0.1487, simple_loss=0.2307, pruned_loss=0.03339, over 7122.00 frames. ], tot_loss[loss=0.1375, simple_loss=0.2179, pruned_loss=0.02852, over 1442273.13 frames. ], batch size: 98, lr: 5.65e-03, grad_scale: 8.0 +2023-03-21 05:23:51,726 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 1.802e+02 2.176e+02 2.517e+02 3.886e+02, threshold=4.352e+02, percent-clipped=0.0 +2023-03-21 05:24:09,866 INFO [train.py:901] (1/2) Epoch 29, batch 2800, loss[loss=0.1211, simple_loss=0.2037, pruned_loss=0.01925, over 7311.00 frames. ], tot_loss[loss=0.1373, simple_loss=0.2179, pruned_loss=0.02833, over 1442725.85 frames. ], batch size: 59, lr: 5.65e-03, grad_scale: 8.0 +2023-03-21 05:24:09,988 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4107, 2.5420, 2.2924, 2.5536, 2.4940, 2.1347, 2.4951, 2.3824], + device='cuda:1'), covar=tensor([0.0610, 0.0544, 0.0818, 0.0635, 0.0630, 0.0704, 0.0711, 0.0802], + device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0053, 0.0061, 0.0053, 0.0051, 0.0055, 0.0053, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:24:34,860 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 05:24:35,195 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 05:24:35,222 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 05:24:35,238 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 05:24:35,406 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 05:24:35,422 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 05:24:35,432 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 05:24:35,528 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 05:24:35,601 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 05:24:35,617 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 05:24:35,633 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 05:24:35,795 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 05:24:35,801 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 05:24:35,808 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 05:24:36,250 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 05:24:36,253 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 05:24:36,309 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 05:24:43,918 INFO [train.py:901] (1/2) Epoch 30, batch 0, loss[loss=0.1409, simple_loss=0.2265, pruned_loss=0.02765, over 7225.00 frames. ], tot_loss[loss=0.1409, simple_loss=0.2265, pruned_loss=0.02765, over 7225.00 frames. ], batch size: 93, lr: 5.56e-03, grad_scale: 8.0 +2023-03-21 05:24:43,919 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 05:24:47,533 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9127, 2.4862, 3.1348, 2.8541, 2.8734, 2.7344, 2.9550, 2.4712], + device='cuda:1'), covar=tensor([0.0296, 0.0326, 0.0588, 0.0485, 0.0575, 0.0808, 0.0367, 0.1760], + device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0344, 0.0274, 0.0362, 0.0301, 0.0301, 0.0349, 0.0271], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:24:56,800 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8342, 2.6268, 2.0520, 3.0722, 1.9339, 2.6103, 1.2767, 2.0555], + device='cuda:1'), covar=tensor([0.0467, 0.0719, 0.2710, 0.0722, 0.0406, 0.0476, 0.3418, 0.1522], + device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0255, 0.0287, 0.0268, 0.0271, 0.0266, 0.0244, 0.0265], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:25:10,054 INFO [train.py:935] (1/2) Epoch 30, validation: loss=0.164, simple_loss=0.254, pruned_loss=0.03703, over 1622729.00 frames. +2023-03-21 05:25:10,054 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 05:25:16,639 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 05:25:17,473 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-03-21 05:25:23,216 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0593, 3.9675, 3.2584, 3.5843, 3.0265, 2.2115, 2.0575, 4.1945], + device='cuda:1'), covar=tensor([0.0059, 0.0111, 0.0134, 0.0061, 0.0143, 0.0527, 0.0544, 0.0044], + device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0087, 0.0105, 0.0090, 0.0121, 0.0128, 0.0125, 0.0097], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 05:25:28,177 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 05:25:29,651 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.214e+02 1.769e+02 2.116e+02 2.355e+02 5.614e+02, threshold=4.232e+02, percent-clipped=2.0 +2023-03-21 05:25:35,182 INFO [train.py:901] (1/2) Epoch 30, batch 50, loss[loss=0.1391, simple_loss=0.2109, pruned_loss=0.03369, over 7348.00 frames. ], tot_loss[loss=0.1372, simple_loss=0.2174, pruned_loss=0.02848, over 325707.36 frames. ], batch size: 51, lr: 5.55e-03, grad_scale: 8.0 +2023-03-21 05:25:35,198 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 05:25:37,228 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 05:25:39,314 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81955.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:25:40,275 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 05:25:40,367 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81957.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:25:56,330 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81987.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:25:57,188 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 05:25:57,753 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 05:26:01,875 INFO [train.py:901] (1/2) Epoch 30, batch 100, loss[loss=0.1375, simple_loss=0.2181, pruned_loss=0.02844, over 7361.00 frames. ], tot_loss[loss=0.1372, simple_loss=0.2175, pruned_loss=0.0285, over 572260.14 frames. ], batch size: 63, lr: 5.55e-03, grad_scale: 8.0 +2023-03-21 05:26:05,354 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=82003.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:26:06,392 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=82005.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:26:16,963 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82026.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 05:26:21,840 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.292e+02 1.747e+02 2.055e+02 2.472e+02 4.255e+02, threshold=4.110e+02, percent-clipped=1.0 +2023-03-21 05:26:27,974 INFO [train.py:901] (1/2) Epoch 30, batch 150, loss[loss=0.1311, simple_loss=0.216, pruned_loss=0.0231, over 7322.00 frames. ], tot_loss[loss=0.1369, simple_loss=0.2176, pruned_loss=0.02811, over 765257.24 frames. ], batch size: 75, lr: 5.55e-03, grad_scale: 8.0 +2023-03-21 05:26:28,653 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82048.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:26:32,204 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82055.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:26:40,398 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 +2023-03-21 05:26:46,244 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2214, 4.6746, 4.7495, 4.6876, 4.6262, 4.2751, 4.7769, 4.6439], + device='cuda:1'), covar=tensor([0.0535, 0.0455, 0.0405, 0.0554, 0.0367, 0.0414, 0.0359, 0.0483], + device='cuda:1'), in_proj_covar=tensor([0.0211, 0.0243, 0.0185, 0.0189, 0.0150, 0.0221, 0.0196, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 05:26:53,597 INFO [train.py:901] (1/2) Epoch 30, batch 200, loss[loss=0.1163, simple_loss=0.1965, pruned_loss=0.01807, over 7133.00 frames. ], tot_loss[loss=0.1359, simple_loss=0.2166, pruned_loss=0.02763, over 917239.77 frames. ], batch size: 41, lr: 5.55e-03, grad_scale: 8.0 +2023-03-21 05:26:58,750 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 05:27:01,864 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3619, 4.3315, 3.9389, 3.8534, 3.4953, 2.5721, 2.1999, 4.4402], + device='cuda:1'), covar=tensor([0.0051, 0.0052, 0.0080, 0.0055, 0.0110, 0.0408, 0.0491, 0.0046], + device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0086, 0.0104, 0.0090, 0.0121, 0.0127, 0.0124, 0.0098], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 05:27:03,305 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 05:27:03,445 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82116.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:27:09,250 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 05:27:13,827 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.287e+02 1.803e+02 2.124e+02 2.586e+02 4.041e+02, threshold=4.248e+02, percent-clipped=0.0 +2023-03-21 05:27:19,424 INFO [train.py:901] (1/2) Epoch 30, batch 250, loss[loss=0.1483, simple_loss=0.2309, pruned_loss=0.03289, over 7345.00 frames. ], tot_loss[loss=0.1365, simple_loss=0.217, pruned_loss=0.02806, over 1034189.04 frames. ], batch size: 51, lr: 5.55e-03, grad_scale: 8.0 +2023-03-21 05:27:22,276 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 05:27:42,103 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 05:27:45,150 INFO [train.py:901] (1/2) Epoch 30, batch 300, loss[loss=0.146, simple_loss=0.2142, pruned_loss=0.03888, over 7263.00 frames. ], tot_loss[loss=0.1377, simple_loss=0.2182, pruned_loss=0.02864, over 1125234.67 frames. ], batch size: 47, lr: 5.55e-03, grad_scale: 8.0 +2023-03-21 05:27:51,575 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 05:28:05,669 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.315e+02 1.824e+02 2.205e+02 2.616e+02 4.930e+02, threshold=4.410e+02, percent-clipped=1.0 +2023-03-21 05:28:11,135 INFO [train.py:901] (1/2) Epoch 30, batch 350, loss[loss=0.1491, simple_loss=0.2344, pruned_loss=0.03191, over 6801.00 frames. ], tot_loss[loss=0.1378, simple_loss=0.2183, pruned_loss=0.02871, over 1195592.54 frames. ], batch size: 107, lr: 5.54e-03, grad_scale: 8.0 +2023-03-21 05:28:25,911 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 05:28:36,973 INFO [train.py:901] (1/2) Epoch 30, batch 400, loss[loss=0.1181, simple_loss=0.1935, pruned_loss=0.02131, over 7129.00 frames. ], tot_loss[loss=0.1375, simple_loss=0.2179, pruned_loss=0.02857, over 1248727.63 frames. ], batch size: 41, lr: 5.54e-03, grad_scale: 8.0 +2023-03-21 05:28:52,522 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82326.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:28:57,388 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.199e+02 1.768e+02 2.131e+02 2.450e+02 4.318e+02, threshold=4.262e+02, percent-clipped=0.0 +2023-03-21 05:29:01,618 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82343.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:29:03,225 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82346.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:29:03,594 INFO [train.py:901] (1/2) Epoch 30, batch 450, loss[loss=0.1475, simple_loss=0.2288, pruned_loss=0.03311, over 7287.00 frames. ], tot_loss[loss=0.1357, simple_loss=0.2158, pruned_loss=0.02776, over 1288248.85 frames. ], batch size: 66, lr: 5.54e-03, grad_scale: 8.0 +2023-03-21 05:29:08,622 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 05:29:08,635 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 05:29:12,757 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7437, 2.9777, 2.5090, 2.9695, 2.8736, 2.5174, 2.9134, 2.7838], + device='cuda:1'), covar=tensor([0.1370, 0.0642, 0.1256, 0.0866, 0.1314, 0.0791, 0.0807, 0.0868], + device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0054, 0.0062, 0.0054, 0.0052, 0.0056, 0.0053, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:29:17,275 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=82374.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:29:29,391 INFO [train.py:901] (1/2) Epoch 30, batch 500, loss[loss=0.1302, simple_loss=0.2146, pruned_loss=0.02294, over 7286.00 frames. ], tot_loss[loss=0.1366, simple_loss=0.217, pruned_loss=0.02805, over 1324574.38 frames. ], batch size: 77, lr: 5.54e-03, grad_scale: 8.0 +2023-03-21 05:29:35,004 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82407.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:29:36,966 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82411.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:29:40,386 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 05:29:42,470 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 05:29:42,972 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 05:29:45,508 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 05:29:50,045 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.309e+02 1.731e+02 2.051e+02 2.399e+02 6.852e+02, threshold=4.103e+02, percent-clipped=1.0 +2023-03-21 05:29:50,069 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 05:29:52,931 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.84 vs. limit=5.0 +2023-03-21 05:29:55,576 INFO [train.py:901] (1/2) Epoch 30, batch 550, loss[loss=0.1252, simple_loss=0.2131, pruned_loss=0.01861, over 7294.00 frames. ], tot_loss[loss=0.1366, simple_loss=0.217, pruned_loss=0.02807, over 1350631.38 frames. ], batch size: 68, lr: 5.54e-03, grad_scale: 8.0 +2023-03-21 05:30:00,947 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 05:30:09,616 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 05:30:13,066 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 05:30:19,377 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82493.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 05:30:19,733 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 05:30:21,206 INFO [train.py:901] (1/2) Epoch 30, batch 600, loss[loss=0.1581, simple_loss=0.2401, pruned_loss=0.03804, over 6706.00 frames. ], tot_loss[loss=0.1371, simple_loss=0.2172, pruned_loss=0.02851, over 1370050.76 frames. ], batch size: 106, lr: 5.54e-03, grad_scale: 8.0 +2023-03-21 05:30:36,840 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 05:30:38,202 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 +2023-03-21 05:30:41,265 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.247e+02 1.785e+02 2.155e+02 2.633e+02 4.871e+02, threshold=4.309e+02, percent-clipped=2.0 +2023-03-21 05:30:45,811 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 05:30:46,856 INFO [train.py:901] (1/2) Epoch 30, batch 650, loss[loss=0.1404, simple_loss=0.2212, pruned_loss=0.02979, over 7329.00 frames. ], tot_loss[loss=0.1369, simple_loss=0.2168, pruned_loss=0.02847, over 1384415.65 frames. ], batch size: 61, lr: 5.53e-03, grad_scale: 8.0 +2023-03-21 05:30:50,542 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82554.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 05:30:51,712 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 +2023-03-21 05:30:58,193 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82569.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:31:03,380 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 05:31:12,418 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 05:31:12,912 INFO [train.py:901] (1/2) Epoch 30, batch 700, loss[loss=0.1289, simple_loss=0.2124, pruned_loss=0.02273, over 7284.00 frames. ], tot_loss[loss=0.1369, simple_loss=0.2168, pruned_loss=0.02852, over 1396479.06 frames. ], batch size: 70, lr: 5.53e-03, grad_scale: 8.0 +2023-03-21 05:31:23,630 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2164, 4.1980, 3.4194, 3.6147, 3.1625, 2.2260, 2.1311, 4.2306], + device='cuda:1'), covar=tensor([0.0046, 0.0043, 0.0098, 0.0065, 0.0120, 0.0484, 0.0490, 0.0039], + device='cuda:1'), in_proj_covar=tensor([0.0104, 0.0085, 0.0103, 0.0089, 0.0120, 0.0126, 0.0123, 0.0097], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 05:31:30,692 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82630.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:31:33,639 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.368e+02 1.896e+02 2.225e+02 2.541e+02 4.418e+02, threshold=4.450e+02, percent-clipped=1.0 +2023-03-21 05:31:36,170 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 05:31:36,681 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 05:31:37,284 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82643.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:31:39,206 INFO [train.py:901] (1/2) Epoch 30, batch 750, loss[loss=0.1409, simple_loss=0.2266, pruned_loss=0.02766, over 7303.00 frames. ], tot_loss[loss=0.1371, simple_loss=0.2172, pruned_loss=0.02854, over 1408453.50 frames. ], batch size: 80, lr: 5.53e-03, grad_scale: 8.0 +2023-03-21 05:31:51,954 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 05:31:52,089 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8234, 2.9905, 2.6468, 2.9756, 3.0075, 2.4679, 2.8461, 2.7957], + device='cuda:1'), covar=tensor([0.0825, 0.0655, 0.0748, 0.0743, 0.0606, 0.0904, 0.1321, 0.1069], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0055, 0.0062, 0.0055, 0.0053, 0.0056, 0.0054, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:31:56,413 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 05:31:56,578 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0991, 2.8820, 2.1254, 3.2811, 2.3816, 2.9258, 1.3837, 1.9948], + device='cuda:1'), covar=tensor([0.0530, 0.0982, 0.2543, 0.0823, 0.0467, 0.0683, 0.3449, 0.1827], + device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0254, 0.0285, 0.0269, 0.0267, 0.0264, 0.0242, 0.0266], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:32:02,519 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=82691.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:32:03,002 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 05:32:04,510 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 05:32:05,526 INFO [train.py:901] (1/2) Epoch 30, batch 800, loss[loss=0.145, simple_loss=0.2237, pruned_loss=0.03315, over 7293.00 frames. ], tot_loss[loss=0.1367, simple_loss=0.2168, pruned_loss=0.02829, over 1416715.54 frames. ], batch size: 66, lr: 5.53e-03, grad_scale: 8.0 +2023-03-21 05:32:06,406 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-21 05:32:08,214 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82702.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:32:09,517 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 +2023-03-21 05:32:12,685 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82711.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:32:15,114 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 05:32:25,181 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.302e+02 1.846e+02 2.089e+02 2.458e+02 4.934e+02, threshold=4.178e+02, percent-clipped=1.0 +2023-03-21 05:32:31,378 INFO [train.py:901] (1/2) Epoch 30, batch 850, loss[loss=0.1463, simple_loss=0.2354, pruned_loss=0.02863, over 6746.00 frames. ], tot_loss[loss=0.1372, simple_loss=0.2172, pruned_loss=0.02857, over 1421938.08 frames. ], batch size: 106, lr: 5.53e-03, grad_scale: 8.0 +2023-03-21 05:32:33,879 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 05:32:34,351 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 05:32:35,984 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6140, 2.7095, 2.4185, 2.7823, 2.7194, 2.3530, 2.7838, 2.5209], + device='cuda:1'), covar=tensor([0.0547, 0.0614, 0.0993, 0.0863, 0.0866, 0.0547, 0.0495, 0.0799], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0054, 0.0062, 0.0055, 0.0053, 0.0056, 0.0053, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:32:37,401 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=82759.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:32:39,927 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 05:32:43,536 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 05:32:47,574 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-21 05:32:57,126 INFO [train.py:901] (1/2) Epoch 30, batch 900, loss[loss=0.1458, simple_loss=0.2303, pruned_loss=0.03059, over 7318.00 frames. ], tot_loss[loss=0.1371, simple_loss=0.2175, pruned_loss=0.02833, over 1428528.00 frames. ], batch size: 59, lr: 5.53e-03, grad_scale: 8.0 +2023-03-21 05:33:04,136 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9887, 2.8734, 3.3299, 2.9318, 3.1153, 2.9275, 2.7430, 3.1967], + device='cuda:1'), covar=tensor([0.2179, 0.0728, 0.0908, 0.1562, 0.1098, 0.1628, 0.1860, 0.1642], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0063, 0.0049, 0.0049, 0.0047, 0.0047, 0.0066, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 05:33:13,290 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-03-21 05:33:17,778 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.226e+02 1.775e+02 2.067e+02 2.438e+02 4.554e+02, threshold=4.134e+02, percent-clipped=2.0 +2023-03-21 05:33:18,611 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-03-21 05:33:23,327 INFO [train.py:901] (1/2) Epoch 30, batch 950, loss[loss=0.1485, simple_loss=0.2354, pruned_loss=0.03077, over 7250.00 frames. ], tot_loss[loss=0.1366, simple_loss=0.217, pruned_loss=0.02815, over 1432993.45 frames. ], batch size: 89, lr: 5.52e-03, grad_scale: 8.0 +2023-03-21 05:33:23,344 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 05:33:24,417 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82849.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 05:33:35,986 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9680, 2.5089, 1.8937, 2.8251, 2.7031, 2.7244, 2.3807, 2.3348], + device='cuda:1'), covar=tensor([0.2284, 0.0954, 0.3535, 0.0660, 0.0263, 0.0197, 0.0340, 0.0367], + device='cuda:1'), in_proj_covar=tensor([0.0259, 0.0234, 0.0261, 0.0264, 0.0191, 0.0186, 0.0209, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 05:33:46,280 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 05:33:49,363 INFO [train.py:901] (1/2) Epoch 30, batch 1000, loss[loss=0.128, simple_loss=0.2166, pruned_loss=0.01973, over 7293.00 frames. ], tot_loss[loss=0.137, simple_loss=0.2174, pruned_loss=0.02828, over 1435915.65 frames. ], batch size: 68, lr: 5.52e-03, grad_scale: 8.0 +2023-03-21 05:34:04,174 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82925.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:34:06,634 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 05:34:08,213 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5721, 4.1024, 4.1778, 4.2080, 4.1816, 4.1331, 4.5113, 4.0865], + device='cuda:1'), covar=tensor([0.0100, 0.0123, 0.0121, 0.0116, 0.0384, 0.0102, 0.0114, 0.0118], + device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0095, 0.0093, 0.0083, 0.0163, 0.0101, 0.0098, 0.0102], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:34:09,599 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 1.830e+02 2.186e+02 2.537e+02 4.453e+02, threshold=4.372e+02, percent-clipped=1.0 +2023-03-21 05:34:15,832 INFO [train.py:901] (1/2) Epoch 30, batch 1050, loss[loss=0.1344, simple_loss=0.2216, pruned_loss=0.02356, over 7222.00 frames. ], tot_loss[loss=0.1369, simple_loss=0.2172, pruned_loss=0.0283, over 1435253.09 frames. ], batch size: 93, lr: 5.52e-03, grad_scale: 8.0 +2023-03-21 05:34:24,064 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82963.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:34:28,458 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 05:34:32,478 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 05:34:41,031 INFO [train.py:901] (1/2) Epoch 30, batch 1100, loss[loss=0.1533, simple_loss=0.2367, pruned_loss=0.03493, over 7285.00 frames. ], tot_loss[loss=0.137, simple_loss=0.2174, pruned_loss=0.02829, over 1438529.76 frames. ], batch size: 70, lr: 5.52e-03, grad_scale: 8.0 +2023-03-21 05:34:44,059 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83002.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:34:55,707 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83024.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:35:00,737 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9624, 4.1090, 3.8998, 4.1205, 3.7498, 4.1120, 4.4404, 4.4218], + device='cuda:1'), covar=tensor([0.0223, 0.0178, 0.0235, 0.0220, 0.0324, 0.0359, 0.0203, 0.0172], + device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0117, 0.0108, 0.0114, 0.0105, 0.0093, 0.0090, 0.0089], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:35:01,610 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.202e+02 1.822e+02 2.194e+02 2.520e+02 4.082e+02, threshold=4.387e+02, percent-clipped=0.0 +2023-03-21 05:35:03,281 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 05:35:03,775 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 05:35:07,767 INFO [train.py:901] (1/2) Epoch 30, batch 1150, loss[loss=0.1147, simple_loss=0.1911, pruned_loss=0.01911, over 6918.00 frames. ], tot_loss[loss=0.1369, simple_loss=0.2173, pruned_loss=0.02828, over 1439499.12 frames. ], batch size: 35, lr: 5.52e-03, grad_scale: 8.0 +2023-03-21 05:35:09,338 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=83050.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:35:16,349 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 05:35:16,846 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 05:35:33,780 INFO [train.py:901] (1/2) Epoch 30, batch 1200, loss[loss=0.1467, simple_loss=0.224, pruned_loss=0.03471, over 7355.00 frames. ], tot_loss[loss=0.1366, simple_loss=0.2169, pruned_loss=0.02817, over 1438245.34 frames. ], batch size: 63, lr: 5.52e-03, grad_scale: 8.0 +2023-03-21 05:35:49,619 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 05:35:50,294 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5022, 1.6681, 1.5223, 1.6227, 1.7004, 1.5144, 1.5165, 1.1715], + device='cuda:1'), covar=tensor([0.0168, 0.0166, 0.0216, 0.0139, 0.0110, 0.0138, 0.0140, 0.0147], + device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0030, 0.0031, 0.0032, 0.0031, 0.0030, 0.0032, 0.0041], + device='cuda:1'), out_proj_covar=tensor([3.8121e-05, 3.4311e-05, 3.4882e-05, 3.5429e-05, 3.4297e-05, 3.3619e-05, + 3.6011e-05, 4.5495e-05], device='cuda:1') +2023-03-21 05:35:54,128 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.348e+02 1.789e+02 2.118e+02 2.602e+02 4.588e+02, threshold=4.235e+02, percent-clipped=1.0 +2023-03-21 05:35:59,540 INFO [train.py:901] (1/2) Epoch 30, batch 1250, loss[loss=0.1504, simple_loss=0.2249, pruned_loss=0.03798, over 7271.00 frames. ], tot_loss[loss=0.1373, simple_loss=0.2176, pruned_loss=0.02849, over 1442047.08 frames. ], batch size: 57, lr: 5.51e-03, grad_scale: 8.0 +2023-03-21 05:36:00,638 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83149.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 05:36:02,641 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0063, 4.1025, 3.4301, 3.6152, 3.0598, 2.2753, 1.9465, 4.0837], + device='cuda:1'), covar=tensor([0.0069, 0.0059, 0.0154, 0.0075, 0.0215, 0.0675, 0.0654, 0.0067], + device='cuda:1'), in_proj_covar=tensor([0.0103, 0.0086, 0.0103, 0.0089, 0.0119, 0.0126, 0.0123, 0.0098], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 05:36:06,753 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6829, 2.5138, 2.3135, 3.7672, 1.7740, 3.4899, 1.3318, 3.2690], + device='cuda:1'), covar=tensor([0.0159, 0.1168, 0.1634, 0.0156, 0.3804, 0.0185, 0.1217, 0.0314], + device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0255, 0.0267, 0.0199, 0.0257, 0.0209, 0.0237, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:36:13,141 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 05:36:13,304 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2684, 2.9134, 1.9550, 3.0011, 3.0278, 3.0848, 2.6455, 2.5153], + device='cuda:1'), covar=tensor([0.2024, 0.0885, 0.3733, 0.0496, 0.0277, 0.0199, 0.0326, 0.0341], + device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0233, 0.0259, 0.0262, 0.0191, 0.0185, 0.0208, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 05:36:17,160 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 05:36:19,315 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 05:36:25,033 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83196.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:36:25,411 INFO [train.py:901] (1/2) Epoch 30, batch 1300, loss[loss=0.1584, simple_loss=0.232, pruned_loss=0.04236, over 7263.00 frames. ], tot_loss[loss=0.1377, simple_loss=0.218, pruned_loss=0.0287, over 1443035.83 frames. ], batch size: 47, lr: 5.51e-03, grad_scale: 8.0 +2023-03-21 05:36:25,470 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=83197.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 05:36:40,696 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83225.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:36:43,088 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 05:36:45,089 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 05:36:46,074 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.311e+02 1.913e+02 2.268e+02 2.716e+02 5.038e+02, threshold=4.536e+02, percent-clipped=3.0 +2023-03-21 05:36:48,609 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 05:36:51,625 INFO [train.py:901] (1/2) Epoch 30, batch 1350, loss[loss=0.1281, simple_loss=0.2173, pruned_loss=0.01949, over 7338.00 frames. ], tot_loss[loss=0.1374, simple_loss=0.2176, pruned_loss=0.02859, over 1443917.16 frames. ], batch size: 44, lr: 5.51e-03, grad_scale: 8.0 +2023-03-21 05:36:56,889 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83257.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:36:59,313 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 05:36:59,444 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5973, 1.3057, 1.6774, 1.9096, 1.7198, 1.9947, 1.5752, 1.9639], + device='cuda:1'), covar=tensor([0.2316, 0.5051, 0.1783, 0.1674, 0.2264, 0.3181, 0.2527, 0.2050], + device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0073, 0.0061, 0.0055, 0.0057, 0.0058, 0.0089, 0.0059], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:37:04,846 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=83273.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:37:11,044 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83284.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:37:17,373 INFO [train.py:901] (1/2) Epoch 30, batch 1400, loss[loss=0.1436, simple_loss=0.224, pruned_loss=0.03155, over 7288.00 frames. ], tot_loss[loss=0.1378, simple_loss=0.218, pruned_loss=0.02879, over 1444187.53 frames. ], batch size: 57, lr: 5.51e-03, grad_scale: 8.0 +2023-03-21 05:37:28,961 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83319.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:37:32,362 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 05:37:33,195 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.83 vs. limit=2.0 +2023-03-21 05:37:37,512 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.214e+02 1.867e+02 2.201e+02 2.472e+02 4.157e+02, threshold=4.401e+02, percent-clipped=0.0 +2023-03-21 05:37:42,155 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83345.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:37:43,000 INFO [train.py:901] (1/2) Epoch 30, batch 1450, loss[loss=0.1328, simple_loss=0.2139, pruned_loss=0.02589, over 7260.00 frames. ], tot_loss[loss=0.1373, simple_loss=0.2177, pruned_loss=0.02841, over 1444862.01 frames. ], batch size: 52, lr: 5.51e-03, grad_scale: 8.0 +2023-03-21 05:37:57,175 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 05:38:09,402 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 05:38:09,521 INFO [train.py:901] (1/2) Epoch 30, batch 1500, loss[loss=0.1269, simple_loss=0.2089, pruned_loss=0.02244, over 7314.00 frames. ], tot_loss[loss=0.137, simple_loss=0.2171, pruned_loss=0.02839, over 1440266.73 frames. ], batch size: 86, lr: 5.51e-03, grad_scale: 8.0 +2023-03-21 05:38:14,472 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 05:38:17,092 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8035, 3.9585, 3.7439, 4.0474, 3.4321, 3.9049, 4.2223, 4.2402], + device='cuda:1'), covar=tensor([0.0242, 0.0176, 0.0232, 0.0188, 0.0415, 0.0337, 0.0292, 0.0232], + device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0119, 0.0110, 0.0116, 0.0107, 0.0095, 0.0092, 0.0091], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:38:29,487 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.731e+02 2.059e+02 2.508e+02 4.024e+02, threshold=4.119e+02, percent-clipped=0.0 +2023-03-21 05:38:34,977 INFO [train.py:901] (1/2) Epoch 30, batch 1550, loss[loss=0.1608, simple_loss=0.24, pruned_loss=0.04086, over 7356.00 frames. ], tot_loss[loss=0.1364, simple_loss=0.2166, pruned_loss=0.02812, over 1440466.25 frames. ], batch size: 63, lr: 5.50e-03, grad_scale: 8.0 +2023-03-21 05:38:39,079 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 05:39:01,218 INFO [train.py:901] (1/2) Epoch 30, batch 1600, loss[loss=0.1445, simple_loss=0.2284, pruned_loss=0.03028, over 7355.00 frames. ], tot_loss[loss=0.1376, simple_loss=0.2177, pruned_loss=0.02879, over 1439356.03 frames. ], batch size: 73, lr: 5.50e-03, grad_scale: 8.0 +2023-03-21 05:39:08,402 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7840, 3.8796, 3.7139, 3.9932, 3.5057, 3.8297, 4.1475, 4.1551], + device='cuda:1'), covar=tensor([0.0255, 0.0194, 0.0225, 0.0186, 0.0411, 0.0313, 0.0266, 0.0227], + device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0120, 0.0110, 0.0116, 0.0107, 0.0095, 0.0093, 0.0091], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:39:10,312 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 05:39:10,793 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 05:39:13,686 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 05:39:19,884 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9035, 2.5091, 3.0021, 2.9217, 3.1027, 2.8703, 2.4890, 2.9460], + device='cuda:1'), covar=tensor([0.1340, 0.0927, 0.1431, 0.1422, 0.0837, 0.0978, 0.2325, 0.1696], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0062, 0.0049, 0.0048, 0.0046, 0.0046, 0.0066, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 05:39:20,731 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.801e+02 2.093e+02 2.671e+02 5.073e+02, threshold=4.186e+02, percent-clipped=2.0 +2023-03-21 05:39:22,803 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 05:39:26,967 INFO [train.py:901] (1/2) Epoch 30, batch 1650, loss[loss=0.1263, simple_loss=0.2091, pruned_loss=0.02176, over 7327.00 frames. ], tot_loss[loss=0.1377, simple_loss=0.2176, pruned_loss=0.02892, over 1439850.19 frames. ], batch size: 61, lr: 5.50e-03, grad_scale: 8.0 +2023-03-21 05:39:28,519 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 05:39:29,621 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83552.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:39:35,727 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3048, 3.2567, 2.4040, 3.8008, 2.8045, 3.3814, 1.5584, 2.3054], + device='cuda:1'), covar=tensor([0.0459, 0.0744, 0.2519, 0.0502, 0.0415, 0.0667, 0.3524, 0.1923], + device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0257, 0.0288, 0.0270, 0.0268, 0.0267, 0.0245, 0.0268], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:39:36,566 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 05:39:39,042 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 +2023-03-21 05:39:40,384 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-03-21 05:39:52,624 INFO [train.py:901] (1/2) Epoch 30, batch 1700, loss[loss=0.1079, simple_loss=0.1764, pruned_loss=0.01973, over 6977.00 frames. ], tot_loss[loss=0.1375, simple_loss=0.2174, pruned_loss=0.02881, over 1440905.55 frames. ], batch size: 35, lr: 5.50e-03, grad_scale: 8.0 +2023-03-21 05:39:54,160 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 05:39:58,540 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 05:40:00,208 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4762, 3.5947, 3.4640, 3.5799, 3.2910, 3.4384, 3.8735, 3.8839], + device='cuda:1'), covar=tensor([0.0257, 0.0197, 0.0244, 0.0211, 0.0365, 0.0389, 0.0242, 0.0198], + device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0118, 0.0109, 0.0115, 0.0106, 0.0094, 0.0092, 0.0090], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:40:04,303 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83619.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:40:04,739 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8667, 4.0086, 3.8525, 4.0318, 3.6175, 3.8726, 4.3126, 4.3209], + device='cuda:1'), covar=tensor([0.0217, 0.0178, 0.0208, 0.0162, 0.0346, 0.0276, 0.0227, 0.0189], + device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0118, 0.0109, 0.0115, 0.0105, 0.0094, 0.0092, 0.0090], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:40:09,660 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 05:40:13,350 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.202e+02 1.754e+02 2.177e+02 2.446e+02 3.446e+02, threshold=4.353e+02, percent-clipped=0.0 +2023-03-21 05:40:15,453 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83640.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:40:18,870 INFO [train.py:901] (1/2) Epoch 30, batch 1750, loss[loss=0.1403, simple_loss=0.2215, pruned_loss=0.02953, over 7352.00 frames. ], tot_loss[loss=0.1371, simple_loss=0.2171, pruned_loss=0.02857, over 1441547.32 frames. ], batch size: 73, lr: 5.50e-03, grad_scale: 8.0 +2023-03-21 05:40:23,587 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1916, 3.3879, 3.9036, 3.8207, 3.4270, 3.3330, 4.2040, 2.9376], + device='cuda:1'), covar=tensor([0.0413, 0.0550, 0.0406, 0.0650, 0.0751, 0.0802, 0.0546, 0.1309], + device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0341, 0.0276, 0.0363, 0.0298, 0.0295, 0.0346, 0.0267], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 05:40:29,504 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=83667.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:40:33,909 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 05:40:34,506 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6375, 3.7565, 3.6188, 3.8021, 3.4086, 3.6408, 4.0158, 4.0759], + device='cuda:1'), covar=tensor([0.0256, 0.0171, 0.0250, 0.0171, 0.0337, 0.0274, 0.0266, 0.0185], + device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0119, 0.0110, 0.0116, 0.0107, 0.0095, 0.0093, 0.0091], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:40:34,937 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 05:40:44,453 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 05:40:44,519 INFO [train.py:901] (1/2) Epoch 30, batch 1800, loss[loss=0.1272, simple_loss=0.2122, pruned_loss=0.02106, over 7287.00 frames. ], tot_loss[loss=0.1371, simple_loss=0.2171, pruned_loss=0.02849, over 1441953.49 frames. ], batch size: 77, lr: 5.50e-03, grad_scale: 16.0 +2023-03-21 05:40:56,073 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 05:40:57,363 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-21 05:41:04,456 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.308e+02 1.893e+02 2.107e+02 2.536e+02 4.167e+02, threshold=4.214e+02, percent-clipped=0.0 +2023-03-21 05:41:09,174 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 05:41:10,634 INFO [train.py:901] (1/2) Epoch 30, batch 1850, loss[loss=0.1432, simple_loss=0.2319, pruned_loss=0.02724, over 7349.00 frames. ], tot_loss[loss=0.1373, simple_loss=0.2176, pruned_loss=0.02846, over 1443620.41 frames. ], batch size: 73, lr: 5.49e-03, grad_scale: 16.0 +2023-03-21 05:41:17,096 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83760.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:41:19,523 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 05:41:32,984 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 +2023-03-21 05:41:35,771 INFO [train.py:901] (1/2) Epoch 30, batch 1900, loss[loss=0.1604, simple_loss=0.2362, pruned_loss=0.04227, over 6687.00 frames. ], tot_loss[loss=0.1373, simple_loss=0.2176, pruned_loss=0.02855, over 1442486.26 frames. ], batch size: 106, lr: 5.49e-03, grad_scale: 16.0 +2023-03-21 05:41:36,774 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 05:41:43,257 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8707, 3.0582, 2.7753, 3.0442, 2.9572, 2.6746, 3.0777, 2.8597], + device='cuda:1'), covar=tensor([0.0471, 0.0856, 0.0771, 0.0900, 0.1570, 0.0746, 0.1036, 0.0967], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0055, 0.0062, 0.0055, 0.0054, 0.0057, 0.0054, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:41:47,300 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1422, 4.2629, 4.0256, 4.2819, 3.8020, 4.1733, 4.5320, 4.5917], + device='cuda:1'), covar=tensor([0.0200, 0.0170, 0.0227, 0.0161, 0.0346, 0.0232, 0.0236, 0.0195], + device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0118, 0.0109, 0.0115, 0.0105, 0.0094, 0.0091, 0.0090], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:41:48,836 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83821.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:41:50,383 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0425, 4.0025, 3.2826, 3.5537, 3.0553, 2.2481, 1.8577, 4.0779], + device='cuda:1'), covar=tensor([0.0042, 0.0052, 0.0120, 0.0063, 0.0141, 0.0535, 0.0609, 0.0048], + device='cuda:1'), in_proj_covar=tensor([0.0104, 0.0086, 0.0104, 0.0089, 0.0120, 0.0128, 0.0124, 0.0098], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 05:41:53,404 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 05:41:56,976 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.794e+02 2.093e+02 2.533e+02 5.661e+02, threshold=4.186e+02, percent-clipped=2.0 +2023-03-21 05:42:02,544 INFO [train.py:901] (1/2) Epoch 30, batch 1950, loss[loss=0.1429, simple_loss=0.2263, pruned_loss=0.02976, over 7310.00 frames. ], tot_loss[loss=0.1371, simple_loss=0.2176, pruned_loss=0.02833, over 1443021.12 frames. ], batch size: 83, lr: 5.49e-03, grad_scale: 16.0 +2023-03-21 05:42:03,057 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 05:42:05,200 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83852.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:42:12,596 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 05:42:17,243 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 05:42:17,736 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 05:42:28,462 INFO [train.py:901] (1/2) Epoch 30, batch 2000, loss[loss=0.1331, simple_loss=0.2183, pruned_loss=0.02394, over 7283.00 frames. ], tot_loss[loss=0.1373, simple_loss=0.2177, pruned_loss=0.02844, over 1445024.49 frames. ], batch size: 68, lr: 5.49e-03, grad_scale: 16.0 +2023-03-21 05:42:30,059 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=83900.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:42:35,465 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 05:42:36,324 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 +2023-03-21 05:42:41,987 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 +2023-03-21 05:42:46,586 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 05:42:48,525 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.334e+02 1.864e+02 2.197e+02 2.640e+02 5.149e+02, threshold=4.395e+02, percent-clipped=2.0 +2023-03-21 05:42:50,676 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83940.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:42:53,969 INFO [train.py:901] (1/2) Epoch 30, batch 2050, loss[loss=0.1471, simple_loss=0.2266, pruned_loss=0.03379, over 7311.00 frames. ], tot_loss[loss=0.1378, simple_loss=0.2182, pruned_loss=0.02868, over 1444156.39 frames. ], batch size: 59, lr: 5.49e-03, grad_scale: 16.0 +2023-03-21 05:42:54,492 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 05:42:55,087 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83949.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:42:59,773 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-21 05:43:13,705 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83985.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:43:15,120 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=83988.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:43:15,674 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0127, 3.7802, 3.7023, 3.6707, 3.6427, 3.5713, 3.8269, 3.5014], + device='cuda:1'), covar=tensor([0.0127, 0.0157, 0.0124, 0.0174, 0.0441, 0.0128, 0.0175, 0.0163], + device='cuda:1'), in_proj_covar=tensor([0.0089, 0.0097, 0.0095, 0.0085, 0.0168, 0.0104, 0.0102, 0.0106], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:43:19,521 INFO [train.py:901] (1/2) Epoch 30, batch 2100, loss[loss=0.1419, simple_loss=0.2251, pruned_loss=0.02937, over 7266.00 frames. ], tot_loss[loss=0.1372, simple_loss=0.2179, pruned_loss=0.02825, over 1444465.06 frames. ], batch size: 47, lr: 5.49e-03, grad_scale: 16.0 +2023-03-21 05:43:29,895 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4209, 2.5214, 2.1134, 3.5107, 1.8964, 3.2180, 1.4206, 3.0221], + device='cuda:1'), covar=tensor([0.0186, 0.1215, 0.1832, 0.0241, 0.3576, 0.0219, 0.1216, 0.0346], + device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0256, 0.0266, 0.0202, 0.0255, 0.0211, 0.0236, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:43:30,887 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84010.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 05:43:32,802 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 05:43:35,799 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 05:43:43,748 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.225e+02 1.924e+02 2.210e+02 2.575e+02 6.608e+02, threshold=4.420e+02, percent-clipped=1.0 +2023-03-21 05:43:48,947 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84046.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:43:49,287 INFO [train.py:901] (1/2) Epoch 30, batch 2150, loss[loss=0.1392, simple_loss=0.2242, pruned_loss=0.02711, over 7289.00 frames. ], tot_loss[loss=0.1374, simple_loss=0.2182, pruned_loss=0.02829, over 1445767.38 frames. ], batch size: 86, lr: 5.48e-03, grad_scale: 16.0 +2023-03-21 05:43:59,331 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3091, 4.8180, 4.9540, 4.8724, 4.8221, 4.3526, 4.9671, 4.8162], + device='cuda:1'), covar=tensor([0.0459, 0.0391, 0.0329, 0.0424, 0.0289, 0.0364, 0.0281, 0.0382], + device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0242, 0.0183, 0.0188, 0.0148, 0.0219, 0.0193, 0.0141], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 05:44:15,531 INFO [train.py:901] (1/2) Epoch 30, batch 2200, loss[loss=0.1377, simple_loss=0.2188, pruned_loss=0.02837, over 7306.00 frames. ], tot_loss[loss=0.1373, simple_loss=0.2182, pruned_loss=0.02824, over 1445321.25 frames. ], batch size: 49, lr: 5.48e-03, grad_scale: 16.0 +2023-03-21 05:44:15,651 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84097.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:44:21,059 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 05:44:25,079 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84116.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:44:34,902 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.268e+02 1.821e+02 2.183e+02 2.536e+02 5.271e+02, threshold=4.365e+02, percent-clipped=2.0 +2023-03-21 05:44:40,502 INFO [train.py:901] (1/2) Epoch 30, batch 2250, loss[loss=0.1468, simple_loss=0.2326, pruned_loss=0.03049, over 6776.00 frames. ], tot_loss[loss=0.1377, simple_loss=0.2187, pruned_loss=0.02837, over 1445291.94 frames. ], batch size: 107, lr: 5.48e-03, grad_scale: 16.0 +2023-03-21 05:44:42,591 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6340, 5.0949, 5.0654, 5.0701, 4.7853, 4.6418, 5.1422, 4.8362], + device='cuda:1'), covar=tensor([0.0804, 0.0790, 0.0743, 0.0846, 0.0766, 0.0684, 0.0647, 0.0985], + device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0246, 0.0185, 0.0191, 0.0149, 0.0222, 0.0196, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 05:44:46,721 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84158.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:44:47,247 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5069, 2.7030, 2.2860, 3.6943, 2.0658, 3.5541, 1.4520, 3.1241], + device='cuda:1'), covar=tensor([0.0117, 0.1107, 0.1642, 0.0158, 0.3399, 0.0184, 0.1211, 0.0274], + device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0256, 0.0269, 0.0203, 0.0258, 0.0212, 0.0238, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:44:54,826 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 05:44:55,313 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 05:44:56,787 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 +2023-03-21 05:45:02,925 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3049, 3.8090, 3.8258, 4.2869, 4.1271, 4.1863, 3.7238, 3.8019], + device='cuda:1'), covar=tensor([0.1068, 0.2971, 0.2567, 0.1226, 0.1160, 0.1602, 0.0991, 0.1530], + device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0380, 0.0287, 0.0296, 0.0218, 0.0361, 0.0216, 0.0262], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:45:06,327 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 05:45:06,653 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.77 vs. limit=5.0 +2023-03-21 05:45:06,805 INFO [train.py:901] (1/2) Epoch 30, batch 2300, loss[loss=0.135, simple_loss=0.2165, pruned_loss=0.02672, over 7263.00 frames. ], tot_loss[loss=0.1381, simple_loss=0.2192, pruned_loss=0.02853, over 1444124.35 frames. ], batch size: 64, lr: 5.48e-03, grad_scale: 16.0 +2023-03-21 05:45:07,979 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84199.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:45:26,699 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.187e+02 1.852e+02 2.178e+02 2.778e+02 7.190e+02, threshold=4.356e+02, percent-clipped=4.0 +2023-03-21 05:45:26,868 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84236.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:45:27,410 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2002, 2.1868, 2.1579, 3.4026, 1.7341, 3.1219, 1.2479, 3.0345], + device='cuda:1'), covar=tensor([0.0181, 0.1477, 0.1834, 0.0228, 0.3956, 0.0207, 0.1323, 0.0413], + device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0257, 0.0269, 0.0204, 0.0259, 0.0212, 0.0240, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:45:32,861 INFO [train.py:901] (1/2) Epoch 30, batch 2350, loss[loss=0.1477, simple_loss=0.2197, pruned_loss=0.03787, over 7215.00 frames. ], tot_loss[loss=0.137, simple_loss=0.2177, pruned_loss=0.02812, over 1440376.22 frames. ], batch size: 45, lr: 5.48e-03, grad_scale: 16.0 +2023-03-21 05:45:35,565 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9358, 3.2153, 3.8153, 3.9791, 4.0118, 3.9948, 3.9655, 3.8801], + device='cuda:1'), covar=tensor([0.0027, 0.0109, 0.0032, 0.0028, 0.0024, 0.0029, 0.0034, 0.0042], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0064, 0.0053, 0.0053, 0.0051, 0.0057, 0.0046, 0.0070], + device='cuda:1'), out_proj_covar=tensor([8.0416e-05, 1.3663e-04, 1.0304e-04, 9.6699e-05, 9.2351e-05, 1.0395e-04, + 9.6302e-05, 1.3563e-04], device='cuda:1') +2023-03-21 05:45:40,170 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84260.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:45:47,701 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0773, 3.3346, 2.9443, 2.9573, 3.0862, 2.7985, 3.1764, 3.0637], + device='cuda:1'), covar=tensor([0.0744, 0.0823, 0.1161, 0.1881, 0.1705, 0.0841, 0.1227, 0.1417], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0055, 0.0063, 0.0054, 0.0053, 0.0056, 0.0054, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:45:50,316 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2973, 2.9763, 3.2258, 3.1306, 2.7969, 2.7283, 3.2424, 2.4233], + device='cuda:1'), covar=tensor([0.0379, 0.0485, 0.0563, 0.0582, 0.0629, 0.0786, 0.0477, 0.1781], + device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0340, 0.0273, 0.0361, 0.0296, 0.0294, 0.0346, 0.0265], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 05:45:54,606 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 05:45:58,581 INFO [train.py:901] (1/2) Epoch 30, batch 2400, loss[loss=0.1335, simple_loss=0.2257, pruned_loss=0.02067, over 7278.00 frames. ], tot_loss[loss=0.137, simple_loss=0.218, pruned_loss=0.02803, over 1440338.38 frames. ], batch size: 70, lr: 5.48e-03, grad_scale: 8.0 +2023-03-21 05:45:58,752 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84297.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:46:00,690 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 05:46:02,718 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84305.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 05:46:06,409 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0388, 2.9413, 2.1544, 3.3708, 2.5933, 3.0605, 1.3913, 2.0632], + device='cuda:1'), covar=tensor([0.0506, 0.0708, 0.2555, 0.0557, 0.0471, 0.0503, 0.3374, 0.1957], + device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0255, 0.0289, 0.0269, 0.0271, 0.0267, 0.0245, 0.0268], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:46:10,763 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 05:46:13,788 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 05:46:18,990 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84336.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:46:19,341 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 1.768e+02 2.044e+02 2.372e+02 6.082e+02, threshold=4.088e+02, percent-clipped=2.0 +2023-03-21 05:46:21,389 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84341.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:46:24,644 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1762, 2.7633, 1.9743, 3.1785, 2.7392, 3.3750, 2.7796, 2.8278], + device='cuda:1'), covar=tensor([0.1775, 0.0867, 0.3377, 0.0564, 0.0201, 0.0212, 0.0280, 0.0333], + device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0233, 0.0256, 0.0261, 0.0190, 0.0187, 0.0208, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 05:46:24,976 INFO [train.py:901] (1/2) Epoch 30, batch 2450, loss[loss=0.1452, simple_loss=0.2255, pruned_loss=0.03242, over 7261.00 frames. ], tot_loss[loss=0.1367, simple_loss=0.2176, pruned_loss=0.02788, over 1442594.94 frames. ], batch size: 55, lr: 5.47e-03, grad_scale: 8.0 +2023-03-21 05:46:40,407 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 05:46:45,116 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9003, 2.2597, 1.7947, 2.7729, 2.5920, 2.7866, 2.2712, 2.5552], + device='cuda:1'), covar=tensor([0.1630, 0.0865, 0.3211, 0.0721, 0.0312, 0.0299, 0.0327, 0.0382], + device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0232, 0.0256, 0.0261, 0.0190, 0.0188, 0.0208, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 05:46:49,910 INFO [train.py:901] (1/2) Epoch 30, batch 2500, loss[loss=0.1314, simple_loss=0.215, pruned_loss=0.02389, over 7269.00 frames. ], tot_loss[loss=0.1373, simple_loss=0.218, pruned_loss=0.02825, over 1442680.87 frames. ], batch size: 77, lr: 5.47e-03, grad_scale: 8.0 +2023-03-21 05:46:50,068 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84397.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:46:52,655 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 +2023-03-21 05:46:59,897 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84416.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:47:06,532 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 05:47:07,244 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0538, 2.7163, 2.8276, 2.9871, 2.5533, 2.5290, 3.0242, 2.1797], + device='cuda:1'), covar=tensor([0.0742, 0.0656, 0.0690, 0.0686, 0.0617, 0.0875, 0.0697, 0.1950], + device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0337, 0.0270, 0.0357, 0.0292, 0.0291, 0.0341, 0.0263], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 05:47:09,658 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4992, 1.7331, 1.4821, 1.6846, 1.8658, 1.4960, 1.5354, 1.3243], + device='cuda:1'), covar=tensor([0.0106, 0.0162, 0.0243, 0.0114, 0.0104, 0.0205, 0.0137, 0.0163], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0031, 0.0030, 0.0031, 0.0031, 0.0029, 0.0032, 0.0041], + device='cuda:1'), out_proj_covar=tensor([3.8375e-05, 3.4443e-05, 3.4406e-05, 3.5121e-05, 3.4294e-05, 3.2850e-05, + 3.6287e-05, 4.6653e-05], device='cuda:1') +2023-03-21 05:47:11,412 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.270e+02 1.780e+02 2.087e+02 2.604e+02 4.810e+02, threshold=4.173e+02, percent-clipped=3.0 +2023-03-21 05:47:16,442 INFO [train.py:901] (1/2) Epoch 30, batch 2550, loss[loss=0.1414, simple_loss=0.2274, pruned_loss=0.02771, over 7328.00 frames. ], tot_loss[loss=0.1378, simple_loss=0.2186, pruned_loss=0.02848, over 1444179.80 frames. ], batch size: 75, lr: 5.47e-03, grad_scale: 8.0 +2023-03-21 05:47:19,552 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84453.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:47:25,098 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=84464.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:47:41,566 INFO [train.py:901] (1/2) Epoch 30, batch 2600, loss[loss=0.131, simple_loss=0.2043, pruned_loss=0.02882, over 7177.00 frames. ], tot_loss[loss=0.1382, simple_loss=0.2189, pruned_loss=0.02873, over 1444275.43 frames. ], batch size: 39, lr: 5.47e-03, grad_scale: 8.0 +2023-03-21 05:47:55,306 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84525.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:47:57,811 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5052, 1.3221, 1.6711, 1.9343, 1.6659, 1.8384, 1.4062, 1.9280], + device='cuda:1'), covar=tensor([0.1872, 0.3371, 0.1449, 0.1477, 0.1094, 0.1313, 0.1707, 0.1188], + device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0074, 0.0062, 0.0057, 0.0057, 0.0060, 0.0093, 0.0059], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 05:48:01,088 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.046e+02 1.691e+02 2.069e+02 2.465e+02 5.601e+02, threshold=4.139e+02, percent-clipped=3.0 +2023-03-21 05:48:06,547 INFO [train.py:901] (1/2) Epoch 30, batch 2650, loss[loss=0.1086, simple_loss=0.1841, pruned_loss=0.01653, over 7204.00 frames. ], tot_loss[loss=0.1377, simple_loss=0.2185, pruned_loss=0.02843, over 1444813.86 frames. ], batch size: 39, lr: 5.47e-03, grad_scale: 8.0 +2023-03-21 05:48:11,013 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84555.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:48:26,244 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84586.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:48:29,149 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84592.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:48:31,474 INFO [train.py:901] (1/2) Epoch 30, batch 2700, loss[loss=0.1416, simple_loss=0.2281, pruned_loss=0.02751, over 7224.00 frames. ], tot_loss[loss=0.1378, simple_loss=0.2187, pruned_loss=0.02844, over 1445783.34 frames. ], batch size: 93, lr: 5.47e-03, grad_scale: 8.0 +2023-03-21 05:48:35,961 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84605.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:48:49,577 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84633.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:48:51,376 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.327e+02 1.736e+02 2.051e+02 2.342e+02 3.644e+02, threshold=4.103e+02, percent-clipped=0.0 +2023-03-21 05:48:53,523 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84641.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:48:54,045 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4477, 2.5653, 2.3688, 3.7548, 1.7609, 3.5039, 1.3711, 3.0332], + device='cuda:1'), covar=tensor([0.0147, 0.1147, 0.1742, 0.0188, 0.3962, 0.0242, 0.1298, 0.0534], + device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0252, 0.0263, 0.0200, 0.0253, 0.0208, 0.0234, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:48:56,450 INFO [train.py:901] (1/2) Epoch 30, batch 2750, loss[loss=0.1276, simple_loss=0.209, pruned_loss=0.02309, over 7324.00 frames. ], tot_loss[loss=0.1375, simple_loss=0.2186, pruned_loss=0.02817, over 1446263.75 frames. ], batch size: 75, lr: 5.47e-03, grad_scale: 8.0 +2023-03-21 05:48:59,520 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=84653.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:49:17,183 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=84689.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:49:18,624 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84692.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:49:19,653 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84694.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:49:21,001 INFO [train.py:901] (1/2) Epoch 30, batch 2800, loss[loss=0.1625, simple_loss=0.2476, pruned_loss=0.03872, over 6727.00 frames. ], tot_loss[loss=0.137, simple_loss=0.2178, pruned_loss=0.02808, over 1441767.24 frames. ], batch size: 106, lr: 5.46e-03, grad_scale: 8.0 +2023-03-21 05:49:45,421 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 05:49:45,483 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 05:49:45,508 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 05:49:45,527 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 05:49:45,699 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 05:49:45,715 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 05:49:45,725 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 05:49:46,117 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 05:49:46,189 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 05:49:46,205 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 05:49:46,221 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 05:49:46,386 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 05:49:46,391 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 05:49:46,398 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 05:49:46,534 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 05:49:46,537 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 05:49:46,599 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 05:49:55,878 INFO [train.py:901] (1/2) Epoch 31, batch 0, loss[loss=0.1246, simple_loss=0.2113, pruned_loss=0.01899, over 7345.00 frames. ], tot_loss[loss=0.1246, simple_loss=0.2113, pruned_loss=0.01899, over 7345.00 frames. ], batch size: 44, lr: 5.37e-03, grad_scale: 8.0 +2023-03-21 05:49:55,878 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 05:50:18,048 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9003, 3.9212, 3.7779, 3.8415, 3.5578, 3.4679, 4.1378, 2.7538], + device='cuda:1'), covar=tensor([0.0500, 0.0417, 0.0375, 0.0569, 0.0628, 0.0708, 0.0558, 0.1514], + device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0336, 0.0271, 0.0357, 0.0294, 0.0291, 0.0343, 0.0264], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 05:50:21,865 INFO [train.py:935] (1/2) Epoch 31, validation: loss=0.1654, simple_loss=0.2551, pruned_loss=0.03787, over 1622729.00 frames. +2023-03-21 05:50:21,866 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 05:50:23,171 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 +2023-03-21 05:50:27,251 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 +2023-03-21 05:50:28,506 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 05:50:29,975 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 1.866e+02 2.119e+02 2.515e+02 6.246e+02, threshold=4.238e+02, percent-clipped=1.0 +2023-03-21 05:50:38,126 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84753.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:50:39,035 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 05:50:46,719 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 05:50:48,339 INFO [train.py:901] (1/2) Epoch 31, batch 50, loss[loss=0.1009, simple_loss=0.1639, pruned_loss=0.01896, over 6150.00 frames. ], tot_loss[loss=0.1372, simple_loss=0.2163, pruned_loss=0.02908, over 323927.43 frames. ], batch size: 26, lr: 5.37e-03, grad_scale: 8.0 +2023-03-21 05:50:49,806 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 05:50:52,828 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 05:50:55,914 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5180, 2.8224, 2.5177, 3.8491, 1.9510, 3.4283, 1.4459, 3.3254], + device='cuda:1'), covar=tensor([0.0131, 0.1083, 0.1603, 0.0155, 0.3513, 0.0193, 0.1186, 0.0401], + device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0253, 0.0264, 0.0201, 0.0254, 0.0209, 0.0236, 0.0229], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:51:03,222 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84800.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:51:03,695 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=84801.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:51:10,618 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 05:51:11,099 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 05:51:13,500 INFO [train.py:901] (1/2) Epoch 31, batch 100, loss[loss=0.1537, simple_loss=0.2347, pruned_loss=0.03629, over 6632.00 frames. ], tot_loss[loss=0.137, simple_loss=0.2164, pruned_loss=0.02882, over 571217.54 frames. ], batch size: 106, lr: 5.37e-03, grad_scale: 8.0 +2023-03-21 05:51:21,416 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 1.807e+02 2.107e+02 2.444e+02 8.407e+02, threshold=4.214e+02, percent-clipped=1.0 +2023-03-21 05:51:27,581 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1530, 2.6956, 2.5911, 2.3651, 2.1363, 2.3861, 2.0703, 2.2401], + device='cuda:1'), covar=tensor([0.0620, 0.0278, 0.0376, 0.0266, 0.0860, 0.0571, 0.0346, 0.0231], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0033, 0.0035, 0.0033, 0.0031, 0.0032, 0.0036, 0.0036], + device='cuda:1'), out_proj_covar=tensor([8.8815e-05, 8.6483e-05, 8.6862e-05, 8.3360e-05, 8.3356e-05, 8.3404e-05, + 8.9069e-05, 9.1886e-05], device='cuda:1') +2023-03-21 05:51:30,583 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84855.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:51:34,854 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84861.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:51:39,597 INFO [train.py:901] (1/2) Epoch 31, batch 150, loss[loss=0.1144, simple_loss=0.1923, pruned_loss=0.01829, over 7200.00 frames. ], tot_loss[loss=0.1356, simple_loss=0.2155, pruned_loss=0.02785, over 765932.85 frames. ], batch size: 39, lr: 5.37e-03, grad_scale: 8.0 +2023-03-21 05:51:44,719 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84881.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:51:50,314 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84892.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:51:55,835 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=84903.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:52:04,874 INFO [train.py:901] (1/2) Epoch 31, batch 200, loss[loss=0.1484, simple_loss=0.2211, pruned_loss=0.03787, over 7262.00 frames. ], tot_loss[loss=0.1357, simple_loss=0.2161, pruned_loss=0.02771, over 916711.11 frames. ], batch size: 64, lr: 5.37e-03, grad_scale: 8.0 +2023-03-21 05:52:09,355 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 05:52:12,946 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.363e+02 1.822e+02 2.128e+02 2.384e+02 5.145e+02, threshold=4.256e+02, percent-clipped=2.0 +2023-03-21 05:52:14,578 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 05:52:15,113 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=84940.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:52:21,243 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 05:52:21,373 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84951.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:52:31,085 INFO [train.py:901] (1/2) Epoch 31, batch 250, loss[loss=0.1293, simple_loss=0.2104, pruned_loss=0.02415, over 7271.00 frames. ], tot_loss[loss=0.1367, simple_loss=0.2174, pruned_loss=0.02803, over 1034214.30 frames. ], batch size: 47, lr: 5.37e-03, grad_scale: 8.0 +2023-03-21 05:52:34,123 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 05:52:40,321 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84989.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:52:41,816 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84992.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:52:52,129 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85012.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:52:53,012 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 05:52:56,386 INFO [train.py:901] (1/2) Epoch 31, batch 300, loss[loss=0.1353, simple_loss=0.2165, pruned_loss=0.02706, over 7280.00 frames. ], tot_loss[loss=0.1375, simple_loss=0.2182, pruned_loss=0.02838, over 1125838.31 frames. ], batch size: 66, lr: 5.37e-03, grad_scale: 8.0 +2023-03-21 05:53:01,374 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.77 vs. limit=5.0 +2023-03-21 05:53:03,982 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 05:53:05,474 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.264e+02 1.823e+02 2.082e+02 2.425e+02 4.646e+02, threshold=4.164e+02, percent-clipped=1.0 +2023-03-21 05:53:06,964 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=85040.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:53:22,337 INFO [train.py:901] (1/2) Epoch 31, batch 350, loss[loss=0.1452, simple_loss=0.2238, pruned_loss=0.03328, over 7216.00 frames. ], tot_loss[loss=0.1366, simple_loss=0.217, pruned_loss=0.02806, over 1195260.83 frames. ], batch size: 45, lr: 5.36e-03, grad_scale: 8.0 +2023-03-21 05:53:36,924 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 05:53:37,815 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 +2023-03-21 05:53:40,538 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85107.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:53:45,793 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85115.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:53:48,633 INFO [train.py:901] (1/2) Epoch 31, batch 400, loss[loss=0.1168, simple_loss=0.1739, pruned_loss=0.0299, over 5868.00 frames. ], tot_loss[loss=0.1361, simple_loss=0.2164, pruned_loss=0.02787, over 1248788.22 frames. ], batch size: 25, lr: 5.36e-03, grad_scale: 8.0 +2023-03-21 05:53:56,683 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.732e+02 2.014e+02 2.395e+02 5.773e+02, threshold=4.029e+02, percent-clipped=2.0 +2023-03-21 05:54:00,138 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 +2023-03-21 05:54:06,348 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85156.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:54:12,500 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85168.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 05:54:13,876 INFO [train.py:901] (1/2) Epoch 31, batch 450, loss[loss=0.1387, simple_loss=0.2213, pruned_loss=0.02802, over 7281.00 frames. ], tot_loss[loss=0.1362, simple_loss=0.2166, pruned_loss=0.02791, over 1292269.47 frames. ], batch size: 77, lr: 5.36e-03, grad_scale: 8.0 +2023-03-21 05:54:16,521 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85176.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:54:19,062 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85181.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:54:19,490 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 05:54:19,986 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 05:54:21,382 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-03-21 05:54:26,639 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8231, 3.1563, 2.5966, 3.0337, 3.0122, 2.7167, 2.9502, 2.9802], + device='cuda:1'), covar=tensor([0.0898, 0.0548, 0.1793, 0.1575, 0.1506, 0.0969, 0.1224, 0.1454], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0055, 0.0063, 0.0055, 0.0053, 0.0057, 0.0054, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:54:40,655 INFO [train.py:901] (1/2) Epoch 31, batch 500, loss[loss=0.1488, simple_loss=0.2336, pruned_loss=0.03196, over 7371.00 frames. ], tot_loss[loss=0.1362, simple_loss=0.2168, pruned_loss=0.02776, over 1326833.91 frames. ], batch size: 63, lr: 5.36e-03, grad_scale: 8.0 +2023-03-21 05:54:43,385 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6640, 3.2132, 3.5022, 3.7386, 3.1593, 2.9836, 3.5382, 2.7252], + device='cuda:1'), covar=tensor([0.0308, 0.0367, 0.0514, 0.0577, 0.0568, 0.0763, 0.0518, 0.1801], + device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0338, 0.0271, 0.0356, 0.0295, 0.0291, 0.0342, 0.0264], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 05:54:44,823 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=85229.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:54:48,810 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.266e+02 1.776e+02 2.130e+02 2.483e+02 6.645e+02, threshold=4.260e+02, percent-clipped=4.0 +2023-03-21 05:54:52,873 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 05:54:54,827 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 05:54:55,368 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 05:54:57,382 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 05:55:02,381 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 05:55:05,845 INFO [train.py:901] (1/2) Epoch 31, batch 550, loss[loss=0.1332, simple_loss=0.2122, pruned_loss=0.02712, over 7237.00 frames. ], tot_loss[loss=0.1357, simple_loss=0.2162, pruned_loss=0.02756, over 1353048.80 frames. ], batch size: 45, lr: 5.36e-03, grad_scale: 8.0 +2023-03-21 05:55:13,464 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 05:55:16,278 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85289.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:55:22,775 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 05:55:25,298 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85307.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:55:25,883 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2327, 3.1339, 2.4579, 3.7324, 2.7564, 3.0908, 1.6313, 2.4243], + device='cuda:1'), covar=tensor([0.0328, 0.0476, 0.2028, 0.0401, 0.0412, 0.0409, 0.3175, 0.1830], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0248, 0.0282, 0.0266, 0.0264, 0.0263, 0.0240, 0.0263], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:55:26,229 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 05:55:32,082 INFO [train.py:901] (1/2) Epoch 31, batch 600, loss[loss=0.1139, simple_loss=0.196, pruned_loss=0.01586, over 7145.00 frames. ], tot_loss[loss=0.1359, simple_loss=0.2163, pruned_loss=0.02777, over 1372898.44 frames. ], batch size: 41, lr: 5.36e-03, grad_scale: 8.0 +2023-03-21 05:55:32,643 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 05:55:40,078 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.256e+02 1.834e+02 2.062e+02 2.649e+02 3.908e+02, threshold=4.123e+02, percent-clipped=0.0 +2023-03-21 05:55:40,156 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=85337.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:55:47,538 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 05:55:50,573 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4402, 3.9295, 3.9677, 4.1384, 3.9854, 3.8806, 4.2274, 3.7818], + device='cuda:1'), covar=tensor([0.0106, 0.0165, 0.0142, 0.0125, 0.0371, 0.0131, 0.0134, 0.0145], + device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0099, 0.0097, 0.0086, 0.0170, 0.0106, 0.0103, 0.0107], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:55:51,086 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9871, 3.4718, 4.0531, 4.0444, 4.2131, 4.1103, 4.0882, 3.9213], + device='cuda:1'), covar=tensor([0.0036, 0.0121, 0.0037, 0.0043, 0.0034, 0.0035, 0.0040, 0.0059], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0065, 0.0053, 0.0054, 0.0052, 0.0058, 0.0047, 0.0071], + device='cuda:1'), out_proj_covar=tensor([7.9437e-05, 1.3907e-04, 1.0249e-04, 9.8237e-05, 9.3299e-05, 1.0564e-04, + 9.6969e-05, 1.3705e-04], device='cuda:1') +2023-03-21 05:55:57,084 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 05:55:57,579 INFO [train.py:901] (1/2) Epoch 31, batch 650, loss[loss=0.1183, simple_loss=0.199, pruned_loss=0.0188, over 7176.00 frames. ], tot_loss[loss=0.1358, simple_loss=0.2163, pruned_loss=0.02765, over 1389642.09 frames. ], batch size: 39, lr: 5.35e-03, grad_scale: 8.0 +2023-03-21 05:56:14,970 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 05:56:15,656 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7293, 3.0675, 2.6663, 3.8765, 1.9059, 3.6364, 1.7621, 3.4698], + device='cuda:1'), covar=tensor([0.0151, 0.0995, 0.1564, 0.0180, 0.4285, 0.0200, 0.1128, 0.0570], + device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0253, 0.0266, 0.0202, 0.0256, 0.0208, 0.0236, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:56:23,514 INFO [train.py:901] (1/2) Epoch 31, batch 700, loss[loss=0.1234, simple_loss=0.2029, pruned_loss=0.02195, over 7339.00 frames. ], tot_loss[loss=0.1359, simple_loss=0.2163, pruned_loss=0.02768, over 1402119.95 frames. ], batch size: 44, lr: 5.35e-03, grad_scale: 8.0 +2023-03-21 05:56:23,548 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 05:56:27,249 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85428.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:56:31,666 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.177e+02 1.664e+02 1.929e+02 2.389e+02 3.949e+02, threshold=3.858e+02, percent-clipped=0.0 +2023-03-21 05:56:31,900 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0542, 2.7414, 3.0348, 3.0843, 2.5937, 2.5488, 3.0474, 2.3851], + device='cuda:1'), covar=tensor([0.0514, 0.0474, 0.0589, 0.0650, 0.0613, 0.0808, 0.0597, 0.1714], + device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0339, 0.0271, 0.0355, 0.0296, 0.0290, 0.0344, 0.0265], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 05:56:35,568 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 +2023-03-21 05:56:41,277 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85456.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:56:45,304 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85463.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 05:56:47,837 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 05:56:48,347 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 05:56:48,727 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 +2023-03-21 05:56:49,810 INFO [train.py:901] (1/2) Epoch 31, batch 750, loss[loss=0.1117, simple_loss=0.1822, pruned_loss=0.02059, over 7004.00 frames. ], tot_loss[loss=0.1358, simple_loss=0.2164, pruned_loss=0.02758, over 1413574.92 frames. ], batch size: 35, lr: 5.35e-03, grad_scale: 8.0 +2023-03-21 05:56:50,018 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85471.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:56:56,115 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85483.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:56:59,211 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85489.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:57:03,245 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 05:57:06,813 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=85504.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:57:07,714 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 05:57:13,599 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 05:57:14,623 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 05:57:15,105 INFO [train.py:901] (1/2) Epoch 31, batch 800, loss[loss=0.119, simple_loss=0.1875, pruned_loss=0.02521, over 7032.00 frames. ], tot_loss[loss=0.1361, simple_loss=0.2164, pruned_loss=0.02783, over 1416291.98 frames. ], batch size: 35, lr: 5.35e-03, grad_scale: 8.0 +2023-03-21 05:57:19,956 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85530.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:57:23,322 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.239e+02 1.779e+02 2.084e+02 2.516e+02 3.956e+02, threshold=4.169e+02, percent-clipped=1.0 +2023-03-21 05:57:24,821 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 05:57:24,923 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9906, 3.4938, 3.9909, 3.8824, 4.0473, 3.9700, 4.0278, 3.8280], + device='cuda:1'), covar=tensor([0.0028, 0.0090, 0.0026, 0.0033, 0.0028, 0.0036, 0.0041, 0.0051], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0065, 0.0053, 0.0053, 0.0052, 0.0057, 0.0047, 0.0071], + device='cuda:1'), out_proj_covar=tensor([7.8922e-05, 1.3826e-04, 1.0134e-04, 9.7635e-05, 9.2574e-05, 1.0496e-04, + 9.5478e-05, 1.3618e-04], device='cuda:1') +2023-03-21 05:57:26,989 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85544.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:57:31,326 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-03-21 05:57:41,519 INFO [train.py:901] (1/2) Epoch 31, batch 850, loss[loss=0.1296, simple_loss=0.2151, pruned_loss=0.02207, over 7323.00 frames. ], tot_loss[loss=0.1361, simple_loss=0.2168, pruned_loss=0.0277, over 1423798.11 frames. ], batch size: 61, lr: 5.35e-03, grad_scale: 8.0 +2023-03-21 05:57:45,032 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 05:57:45,041 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 05:57:49,988 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 +2023-03-21 05:57:50,155 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 05:57:51,812 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85591.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:57:53,767 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 05:58:00,190 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85607.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:58:02,284 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5159, 1.5650, 1.4598, 1.4922, 1.6979, 1.4763, 1.5194, 1.0292], + device='cuda:1'), covar=tensor([0.0174, 0.0140, 0.0344, 0.0212, 0.0126, 0.0133, 0.0133, 0.0194], + device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0029, 0.0029, 0.0031, 0.0029, 0.0028, 0.0031, 0.0040], + device='cuda:1'), out_proj_covar=tensor([3.6833e-05, 3.3250e-05, 3.2855e-05, 3.4117e-05, 3.2290e-05, 3.1624e-05, + 3.5009e-05, 4.4660e-05], device='cuda:1') +2023-03-21 05:58:02,312 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3995, 3.3004, 2.2433, 3.6236, 2.6782, 3.2963, 1.5299, 2.3280], + device='cuda:1'), covar=tensor([0.0501, 0.1013, 0.2637, 0.0769, 0.0503, 0.0790, 0.3846, 0.1856], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0250, 0.0282, 0.0267, 0.0264, 0.0264, 0.0240, 0.0261], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 05:58:07,162 INFO [train.py:901] (1/2) Epoch 31, batch 900, loss[loss=0.1415, simple_loss=0.2235, pruned_loss=0.0297, over 7262.00 frames. ], tot_loss[loss=0.1366, simple_loss=0.2169, pruned_loss=0.02813, over 1427177.98 frames. ], batch size: 89, lr: 5.35e-03, grad_scale: 8.0 +2023-03-21 05:58:09,305 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85625.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:58:15,249 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.228e+02 1.810e+02 2.129e+02 2.402e+02 5.713e+02, threshold=4.258e+02, percent-clipped=2.0 +2023-03-21 05:58:15,440 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8731, 3.1934, 2.6761, 3.1228, 3.0350, 2.6325, 3.0063, 3.0463], + device='cuda:1'), covar=tensor([0.0887, 0.0725, 0.1226, 0.0867, 0.0854, 0.0667, 0.1070, 0.0897], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0056, 0.0064, 0.0055, 0.0054, 0.0057, 0.0055, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 05:58:25,755 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=85655.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:58:33,620 INFO [train.py:901] (1/2) Epoch 31, batch 950, loss[loss=0.1325, simple_loss=0.2127, pruned_loss=0.02611, over 7271.00 frames. ], tot_loss[loss=0.1359, simple_loss=0.2158, pruned_loss=0.02805, over 1425947.07 frames. ], batch size: 47, lr: 5.34e-03, grad_scale: 8.0 +2023-03-21 05:58:34,625 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 05:58:41,343 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85686.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 05:58:58,445 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 05:58:58,920 INFO [train.py:901] (1/2) Epoch 31, batch 1000, loss[loss=0.1607, simple_loss=0.2484, pruned_loss=0.03646, over 6721.00 frames. ], tot_loss[loss=0.1361, simple_loss=0.216, pruned_loss=0.02812, over 1428814.19 frames. ], batch size: 107, lr: 5.34e-03, grad_scale: 8.0 +2023-03-21 05:59:08,124 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.756e+02 2.066e+02 2.390e+02 3.507e+02, threshold=4.132e+02, percent-clipped=0.0 +2023-03-21 05:59:19,083 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 05:59:19,709 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7996, 1.7591, 2.0986, 2.5391, 2.1657, 2.4519, 2.2434, 2.3702], + device='cuda:1'), covar=tensor([0.3230, 0.3018, 0.1974, 0.0879, 0.3598, 0.1681, 0.1724, 0.2267], + device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0072, 0.0062, 0.0057, 0.0057, 0.0057, 0.0093, 0.0059], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 05:59:21,196 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85763.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:59:25,128 INFO [train.py:901] (1/2) Epoch 31, batch 1050, loss[loss=0.1226, simple_loss=0.1987, pruned_loss=0.02327, over 7253.00 frames. ], tot_loss[loss=0.1362, simple_loss=0.2162, pruned_loss=0.02809, over 1432559.34 frames. ], batch size: 47, lr: 5.34e-03, grad_scale: 8.0 +2023-03-21 05:59:25,254 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85771.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:59:31,709 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85784.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:59:39,674 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 05:59:43,958 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 05:59:45,479 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=85811.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:59:50,040 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=85819.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 05:59:50,974 INFO [train.py:901] (1/2) Epoch 31, batch 1100, loss[loss=0.1271, simple_loss=0.2117, pruned_loss=0.02125, over 7276.00 frames. ], tot_loss[loss=0.1366, simple_loss=0.2168, pruned_loss=0.0282, over 1435188.62 frames. ], batch size: 66, lr: 5.34e-03, grad_scale: 8.0 +2023-03-21 05:59:55,275 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0464, 2.4999, 1.9256, 2.7765, 2.8867, 3.1144, 2.4489, 2.4187], + device='cuda:1'), covar=tensor([0.1708, 0.0893, 0.3132, 0.0662, 0.0231, 0.0182, 0.0273, 0.0311], + device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0233, 0.0252, 0.0260, 0.0188, 0.0186, 0.0209, 0.0223], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 05:59:59,600 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.280e+02 1.827e+02 2.007e+02 2.454e+02 6.338e+02, threshold=4.013e+02, percent-clipped=1.0 +2023-03-21 06:00:00,720 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85839.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:00:13,667 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 06:00:14,164 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 06:00:16,637 INFO [train.py:901] (1/2) Epoch 31, batch 1150, loss[loss=0.129, simple_loss=0.2087, pruned_loss=0.02461, over 7327.00 frames. ], tot_loss[loss=0.1364, simple_loss=0.2167, pruned_loss=0.02806, over 1437529.23 frames. ], batch size: 83, lr: 5.34e-03, grad_scale: 8.0 +2023-03-21 06:00:18,811 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85875.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:00:24,189 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85886.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:00:25,675 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 06:00:26,687 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 06:00:42,910 INFO [train.py:901] (1/2) Epoch 31, batch 1200, loss[loss=0.1107, simple_loss=0.1711, pruned_loss=0.02516, over 6297.00 frames. ], tot_loss[loss=0.1364, simple_loss=0.2165, pruned_loss=0.02815, over 1434226.16 frames. ], batch size: 27, lr: 5.34e-03, grad_scale: 8.0 +2023-03-21 06:00:47,580 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85930.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:00:50,542 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85936.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:00:50,864 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.241e+02 1.650e+02 2.176e+02 2.512e+02 6.235e+02, threshold=4.352e+02, percent-clipped=4.0 +2023-03-21 06:00:58,476 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 06:01:01,067 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85957.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:01:07,949 INFO [train.py:901] (1/2) Epoch 31, batch 1250, loss[loss=0.1323, simple_loss=0.2109, pruned_loss=0.02679, over 7301.00 frames. ], tot_loss[loss=0.1364, simple_loss=0.2166, pruned_loss=0.02804, over 1436197.07 frames. ], batch size: 80, lr: 5.34e-03, grad_scale: 8.0 +2023-03-21 06:01:13,064 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85981.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 06:01:18,834 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85991.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 06:01:22,810 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 06:01:25,295 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86002.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:01:25,974 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 +2023-03-21 06:01:26,717 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 06:01:27,707 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 06:01:33,423 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86018.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:01:34,748 INFO [train.py:901] (1/2) Epoch 31, batch 1300, loss[loss=0.1312, simple_loss=0.2106, pruned_loss=0.02583, over 7161.00 frames. ], tot_loss[loss=0.1367, simple_loss=0.217, pruned_loss=0.02816, over 1438122.92 frames. ], batch size: 41, lr: 5.33e-03, grad_scale: 8.0 +2023-03-21 06:01:42,743 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.255e+02 1.873e+02 2.179e+02 2.482e+02 5.073e+02, threshold=4.358e+02, percent-clipped=1.0 +2023-03-21 06:01:46,682 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 +2023-03-21 06:01:50,400 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86052.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:01:51,265 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 06:01:53,815 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 06:01:55,945 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86063.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:01:56,794 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 06:01:58,931 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86069.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:01:59,809 INFO [train.py:901] (1/2) Epoch 31, batch 1350, loss[loss=0.1274, simple_loss=0.2119, pruned_loss=0.02146, over 7228.00 frames. ], tot_loss[loss=0.1374, simple_loss=0.2178, pruned_loss=0.02845, over 1439444.39 frames. ], batch size: 45, lr: 5.33e-03, grad_scale: 8.0 +2023-03-21 06:02:07,182 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86084.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:02:08,676 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 06:02:22,233 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86113.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:02:26,058 INFO [train.py:901] (1/2) Epoch 31, batch 1400, loss[loss=0.1354, simple_loss=0.2206, pruned_loss=0.02514, over 7372.00 frames. ], tot_loss[loss=0.137, simple_loss=0.2175, pruned_loss=0.02823, over 1440752.62 frames. ], batch size: 65, lr: 5.33e-03, grad_scale: 8.0 +2023-03-21 06:02:30,707 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86130.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:02:31,581 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=86132.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:02:34,025 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.336e+02 1.787e+02 2.113e+02 2.434e+02 4.644e+02, threshold=4.226e+02, percent-clipped=1.0 +2023-03-21 06:02:35,162 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86139.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:02:39,701 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 06:02:44,637 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 +2023-03-21 06:02:52,507 INFO [train.py:901] (1/2) Epoch 31, batch 1450, loss[loss=0.1599, simple_loss=0.2411, pruned_loss=0.0394, over 6741.00 frames. ], tot_loss[loss=0.1362, simple_loss=0.2166, pruned_loss=0.02793, over 1435958.40 frames. ], batch size: 107, lr: 5.33e-03, grad_scale: 8.0 +2023-03-21 06:02:56,285 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-03-21 06:03:00,201 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86186.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:03:00,666 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=86187.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:03:01,258 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2745, 3.6951, 3.0792, 3.5509, 3.4747, 2.8345, 3.4740, 3.2849], + device='cuda:1'), covar=tensor([0.0822, 0.1100, 0.0790, 0.0565, 0.0883, 0.0737, 0.0635, 0.1435], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0055, 0.0063, 0.0055, 0.0053, 0.0057, 0.0055, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:03:04,161 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 06:03:18,057 INFO [train.py:901] (1/2) Epoch 31, batch 1500, loss[loss=0.1569, simple_loss=0.2337, pruned_loss=0.04006, over 7298.00 frames. ], tot_loss[loss=0.1363, simple_loss=0.2167, pruned_loss=0.02799, over 1438930.41 frames. ], batch size: 86, lr: 5.33e-03, grad_scale: 8.0 +2023-03-21 06:03:20,508 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 06:03:23,107 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86231.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:03:24,592 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=86234.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:03:25,988 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.325e+02 1.734e+02 2.049e+02 2.482e+02 4.437e+02, threshold=4.097e+02, percent-clipped=1.0 +2023-03-21 06:03:31,409 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86246.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:03:44,518 INFO [train.py:901] (1/2) Epoch 31, batch 1550, loss[loss=0.1311, simple_loss=0.2153, pruned_loss=0.02347, over 7305.00 frames. ], tot_loss[loss=0.1364, simple_loss=0.2168, pruned_loss=0.028, over 1441055.79 frames. ], batch size: 83, lr: 5.33e-03, grad_scale: 8.0 +2023-03-21 06:03:46,056 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 06:03:49,779 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86281.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:03:52,209 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86286.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:04:02,349 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1905, 2.5973, 1.9821, 3.0084, 2.9431, 3.2587, 2.6163, 2.5781], + device='cuda:1'), covar=tensor([0.2077, 0.0968, 0.3568, 0.0472, 0.0229, 0.0233, 0.0401, 0.0389], + device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0237, 0.0259, 0.0265, 0.0193, 0.0192, 0.0215, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 06:04:02,827 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86307.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:04:05,844 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86313.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:04:09,816 INFO [train.py:901] (1/2) Epoch 31, batch 1600, loss[loss=0.1256, simple_loss=0.212, pruned_loss=0.01955, over 7356.00 frames. ], tot_loss[loss=0.1367, simple_loss=0.2171, pruned_loss=0.02812, over 1440868.52 frames. ], batch size: 51, lr: 5.32e-03, grad_scale: 16.0 +2023-03-21 06:04:13,935 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=86329.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:04:16,983 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 06:04:17,497 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 06:04:18,551 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.253e+02 1.767e+02 1.989e+02 2.470e+02 3.737e+02, threshold=3.978e+02, percent-clipped=0.0 +2023-03-21 06:04:20,570 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 06:04:29,950 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86358.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:04:30,890 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 06:04:35,363 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 06:04:36,364 INFO [train.py:901] (1/2) Epoch 31, batch 1650, loss[loss=0.1303, simple_loss=0.2193, pruned_loss=0.02064, over 7252.00 frames. ], tot_loss[loss=0.1359, simple_loss=0.2162, pruned_loss=0.02783, over 1438869.29 frames. ], batch size: 55, lr: 5.32e-03, grad_scale: 16.0 +2023-03-21 06:04:36,552 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3425, 2.6304, 2.9488, 3.0507, 2.7689, 2.7898, 3.1122, 2.3425], + device='cuda:1'), covar=tensor([0.0500, 0.0551, 0.0686, 0.0650, 0.0610, 0.1032, 0.0680, 0.2026], + device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0340, 0.0275, 0.0356, 0.0295, 0.0292, 0.0345, 0.0265], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:04:43,249 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 06:04:55,291 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86408.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:05:00,278 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 06:05:01,541 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1416, 2.6883, 1.9395, 3.0027, 2.8827, 3.0493, 2.6459, 2.7308], + device='cuda:1'), covar=tensor([0.1853, 0.0862, 0.3377, 0.0669, 0.0238, 0.0201, 0.0361, 0.0374], + device='cuda:1'), in_proj_covar=tensor([0.0259, 0.0235, 0.0256, 0.0263, 0.0191, 0.0190, 0.0213, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 06:05:02,422 INFO [train.py:901] (1/2) Epoch 31, batch 1700, loss[loss=0.1225, simple_loss=0.1859, pruned_loss=0.02952, over 6547.00 frames. ], tot_loss[loss=0.1366, simple_loss=0.2167, pruned_loss=0.02824, over 1439665.42 frames. ], batch size: 28, lr: 5.32e-03, grad_scale: 16.0 +2023-03-21 06:05:04,447 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86425.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:05:04,537 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6446, 1.8811, 1.5803, 1.7935, 1.9947, 1.6647, 1.5567, 1.3111], + device='cuda:1'), covar=tensor([0.0133, 0.0155, 0.0220, 0.0125, 0.0084, 0.0112, 0.0228, 0.0172], + device='cuda:1'), in_proj_covar=tensor([0.0032, 0.0029, 0.0029, 0.0030, 0.0029, 0.0028, 0.0031, 0.0040], + device='cuda:1'), out_proj_covar=tensor([3.6867e-05, 3.3223e-05, 3.3459e-05, 3.3785e-05, 3.2984e-05, 3.1536e-05, + 3.5046e-05, 4.5143e-05], device='cuda:1') +2023-03-21 06:05:04,920 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 06:05:11,065 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.287e+02 1.933e+02 2.262e+02 2.614e+02 4.811e+02, threshold=4.525e+02, percent-clipped=2.0 +2023-03-21 06:05:12,779 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86440.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:05:15,316 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86445.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:05:15,665 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 06:05:28,111 INFO [train.py:901] (1/2) Epoch 31, batch 1750, loss[loss=0.1391, simple_loss=0.224, pruned_loss=0.02713, over 7222.00 frames. ], tot_loss[loss=0.1366, simple_loss=0.2168, pruned_loss=0.02824, over 1441959.63 frames. ], batch size: 93, lr: 5.32e-03, grad_scale: 16.0 +2023-03-21 06:05:40,318 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 06:05:41,757 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 06:05:43,416 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86501.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:05:45,927 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86506.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 06:05:53,963 INFO [train.py:901] (1/2) Epoch 31, batch 1800, loss[loss=0.1288, simple_loss=0.2073, pruned_loss=0.02519, over 7358.00 frames. ], tot_loss[loss=0.1367, simple_loss=0.217, pruned_loss=0.02822, over 1441952.06 frames. ], batch size: 73, lr: 5.32e-03, grad_scale: 16.0 +2023-03-21 06:05:59,635 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86531.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:06:02,538 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.341e+02 1.702e+02 2.082e+02 2.428e+02 3.685e+02, threshold=4.165e+02, percent-clipped=0.0 +2023-03-21 06:06:03,550 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 06:06:11,891 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-21 06:06:16,549 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 06:06:16,710 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0471, 2.4903, 1.9254, 2.7447, 2.6967, 2.9254, 2.3830, 2.5015], + device='cuda:1'), covar=tensor([0.2213, 0.1012, 0.3547, 0.0559, 0.0275, 0.0326, 0.0347, 0.0456], + device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0234, 0.0254, 0.0260, 0.0191, 0.0190, 0.0211, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 06:06:19,519 INFO [train.py:901] (1/2) Epoch 31, batch 1850, loss[loss=0.1306, simple_loss=0.2176, pruned_loss=0.02182, over 7332.00 frames. ], tot_loss[loss=0.1367, simple_loss=0.2171, pruned_loss=0.02814, over 1443414.31 frames. ], batch size: 75, lr: 5.32e-03, grad_scale: 16.0 +2023-03-21 06:06:23,588 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=86579.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:06:24,175 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86580.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:06:26,032 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 06:06:27,098 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86586.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 06:06:36,058 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86602.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:06:42,327 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86613.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:06:44,700 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 06:06:45,792 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3070, 4.3229, 3.6140, 3.7686, 3.1448, 2.4683, 1.8976, 4.3094], + device='cuda:1'), covar=tensor([0.0049, 0.0047, 0.0116, 0.0065, 0.0182, 0.0500, 0.0633, 0.0048], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0085, 0.0103, 0.0088, 0.0117, 0.0125, 0.0123, 0.0096], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 06:06:46,165 INFO [train.py:901] (1/2) Epoch 31, batch 1900, loss[loss=0.1586, simple_loss=0.2371, pruned_loss=0.04007, over 7296.00 frames. ], tot_loss[loss=0.1365, simple_loss=0.2172, pruned_loss=0.02787, over 1445408.92 frames. ], batch size: 49, lr: 5.32e-03, grad_scale: 16.0 +2023-03-21 06:06:52,827 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=86634.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:06:54,240 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.252e+02 1.720e+02 1.969e+02 2.412e+02 4.198e+02, threshold=3.938e+02, percent-clipped=1.0 +2023-03-21 06:06:56,450 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86641.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:07:05,047 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86658.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:07:06,540 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=86661.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:07:11,027 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 06:07:11,519 INFO [train.py:901] (1/2) Epoch 31, batch 1950, loss[loss=0.1556, simple_loss=0.2312, pruned_loss=0.03993, over 7277.00 frames. ], tot_loss[loss=0.1361, simple_loss=0.2169, pruned_loss=0.02764, over 1444371.57 frames. ], batch size: 52, lr: 5.31e-03, grad_scale: 16.0 +2023-03-21 06:07:22,669 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 06:07:27,200 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 06:07:27,508 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 +2023-03-21 06:07:27,688 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 06:07:30,446 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=86706.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:07:31,549 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86708.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:07:38,004 INFO [train.py:901] (1/2) Epoch 31, batch 2000, loss[loss=0.1536, simple_loss=0.2368, pruned_loss=0.03519, over 6725.00 frames. ], tot_loss[loss=0.1365, simple_loss=0.2174, pruned_loss=0.0278, over 1444373.12 frames. ], batch size: 107, lr: 5.31e-03, grad_scale: 16.0 +2023-03-21 06:07:40,041 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86725.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:07:43,438 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 06:07:45,925 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 1.843e+02 2.081e+02 2.436e+02 4.240e+02, threshold=4.161e+02, percent-clipped=2.0 +2023-03-21 06:07:54,456 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 06:07:55,481 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=86756.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:07:55,756 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-21 06:08:02,008 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 06:08:03,470 INFO [train.py:901] (1/2) Epoch 31, batch 2050, loss[loss=0.1472, simple_loss=0.2344, pruned_loss=0.03006, over 6701.00 frames. ], tot_loss[loss=0.1375, simple_loss=0.2185, pruned_loss=0.02823, over 1443490.76 frames. ], batch size: 106, lr: 5.31e-03, grad_scale: 16.0 +2023-03-21 06:08:04,512 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=86773.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:08:12,408 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9303, 2.5901, 2.9094, 2.8837, 2.4544, 2.6421, 3.0657, 2.2760], + device='cuda:1'), covar=tensor([0.0557, 0.0523, 0.0602, 0.0673, 0.0600, 0.0881, 0.0727, 0.1881], + device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0335, 0.0271, 0.0353, 0.0291, 0.0290, 0.0343, 0.0263], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:08:14,269 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0866, 4.5322, 4.6064, 4.5391, 4.5429, 4.1024, 4.6359, 4.4899], + device='cuda:1'), covar=tensor([0.0504, 0.0421, 0.0400, 0.0533, 0.0338, 0.0426, 0.0332, 0.0466], + device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0241, 0.0186, 0.0187, 0.0147, 0.0217, 0.0190, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:08:15,791 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0546, 4.3220, 3.9148, 4.3458, 3.8227, 4.2409, 4.5583, 4.6168], + device='cuda:1'), covar=tensor([0.0233, 0.0135, 0.0236, 0.0155, 0.0315, 0.0337, 0.0286, 0.0186], + device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0121, 0.0115, 0.0118, 0.0110, 0.0100, 0.0096, 0.0095], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:08:16,791 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86796.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:08:19,694 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86801.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 06:08:26,253 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86814.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:08:29,727 INFO [train.py:901] (1/2) Epoch 31, batch 2100, loss[loss=0.1333, simple_loss=0.2089, pruned_loss=0.02884, over 7347.00 frames. ], tot_loss[loss=0.1373, simple_loss=0.2181, pruned_loss=0.02824, over 1443496.02 frames. ], batch size: 63, lr: 5.31e-03, grad_scale: 16.0 +2023-03-21 06:08:35,867 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 06:08:36,133 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-03-21 06:08:37,849 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.065e+02 1.745e+02 2.102e+02 2.497e+02 4.075e+02, threshold=4.204e+02, percent-clipped=0.0 +2023-03-21 06:08:38,877 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 06:08:56,246 INFO [train.py:901] (1/2) Epoch 31, batch 2150, loss[loss=0.1131, simple_loss=0.1912, pruned_loss=0.01745, over 7131.00 frames. ], tot_loss[loss=0.1373, simple_loss=0.2181, pruned_loss=0.02823, over 1443675.88 frames. ], batch size: 41, lr: 5.31e-03, grad_scale: 16.0 +2023-03-21 06:08:58,374 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86875.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:08:59,814 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86878.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:09:11,982 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86902.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:09:21,488 INFO [train.py:901] (1/2) Epoch 31, batch 2200, loss[loss=0.1405, simple_loss=0.2135, pruned_loss=0.03371, over 7214.00 frames. ], tot_loss[loss=0.1371, simple_loss=0.2181, pruned_loss=0.02806, over 1444415.86 frames. ], batch size: 45, lr: 5.31e-03, grad_scale: 16.0 +2023-03-21 06:09:23,555 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 06:09:29,172 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86936.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:09:29,576 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.237e+02 1.856e+02 2.260e+02 2.640e+02 6.154e+02, threshold=4.519e+02, percent-clipped=3.0 +2023-03-21 06:09:31,378 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86939.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:09:36,809 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=86950.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:09:44,977 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4977, 4.5627, 4.4108, 4.5989, 4.5286, 4.1917, 4.6048, 4.6775], + device='cuda:1'), covar=tensor([0.0304, 0.0209, 0.0273, 0.0284, 0.0303, 0.0413, 0.0500, 0.0373], + device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0121, 0.0114, 0.0118, 0.0109, 0.0099, 0.0095, 0.0094], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:09:47,939 INFO [train.py:901] (1/2) Epoch 31, batch 2250, loss[loss=0.1319, simple_loss=0.2182, pruned_loss=0.02279, over 7271.00 frames. ], tot_loss[loss=0.1361, simple_loss=0.2171, pruned_loss=0.02758, over 1441542.40 frames. ], batch size: 94, lr: 5.30e-03, grad_scale: 16.0 +2023-03-21 06:09:52,020 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7822, 5.3604, 5.4253, 5.3544, 5.1546, 4.8372, 5.4793, 5.2348], + device='cuda:1'), covar=tensor([0.0509, 0.0356, 0.0381, 0.0473, 0.0315, 0.0386, 0.0284, 0.0419], + device='cuda:1'), in_proj_covar=tensor([0.0216, 0.0249, 0.0193, 0.0193, 0.0151, 0.0224, 0.0195, 0.0146], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:09:58,663 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 06:09:59,136 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 06:10:04,160 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87002.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 06:10:05,827 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.50 vs. limit=5.0 +2023-03-21 06:10:07,827 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.06 vs. limit=5.0 +2023-03-21 06:10:11,491 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 06:10:13,515 INFO [train.py:901] (1/2) Epoch 31, batch 2300, loss[loss=0.1436, simple_loss=0.233, pruned_loss=0.02713, over 7291.00 frames. ], tot_loss[loss=0.1353, simple_loss=0.2161, pruned_loss=0.02722, over 1441770.91 frames. ], batch size: 86, lr: 5.30e-03, grad_scale: 16.0 +2023-03-21 06:10:22,192 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.208e+02 1.765e+02 2.120e+02 2.514e+02 3.861e+02, threshold=4.240e+02, percent-clipped=0.0 +2023-03-21 06:10:36,056 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87063.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 06:10:39,860 INFO [train.py:901] (1/2) Epoch 31, batch 2350, loss[loss=0.1316, simple_loss=0.2187, pruned_loss=0.02227, over 7330.00 frames. ], tot_loss[loss=0.1353, simple_loss=0.2163, pruned_loss=0.02712, over 1441141.89 frames. ], batch size: 83, lr: 5.30e-03, grad_scale: 16.0 +2023-03-21 06:10:40,158 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-03-21 06:10:52,314 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87096.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:10:54,775 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87101.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:10:55,959 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 +2023-03-21 06:10:57,633 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 06:11:03,798 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 06:11:05,267 INFO [train.py:901] (1/2) Epoch 31, batch 2400, loss[loss=0.1406, simple_loss=0.2299, pruned_loss=0.02564, over 6829.00 frames. ], tot_loss[loss=0.1363, simple_loss=0.2176, pruned_loss=0.0275, over 1442600.75 frames. ], batch size: 107, lr: 5.30e-03, grad_scale: 16.0 +2023-03-21 06:11:13,831 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.314e+02 1.833e+02 2.136e+02 2.421e+02 3.611e+02, threshold=4.272e+02, percent-clipped=0.0 +2023-03-21 06:11:14,355 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 06:11:17,464 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 06:11:17,527 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=87144.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:11:20,020 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=87149.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:11:30,745 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87170.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:11:31,139 INFO [train.py:901] (1/2) Epoch 31, batch 2450, loss[loss=0.1427, simple_loss=0.2151, pruned_loss=0.03513, over 7271.00 frames. ], tot_loss[loss=0.1357, simple_loss=0.2168, pruned_loss=0.0273, over 1440154.89 frames. ], batch size: 70, lr: 5.30e-03, grad_scale: 16.0 +2023-03-21 06:11:35,828 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1794, 2.7164, 3.3313, 3.0292, 3.2537, 3.0444, 2.7614, 3.1672], + device='cuda:1'), covar=tensor([0.1616, 0.1133, 0.1210, 0.1622, 0.0818, 0.1094, 0.2133, 0.1775], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0050, 0.0048, 0.0048, 0.0047, 0.0068, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 06:11:43,159 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 06:11:57,939 INFO [train.py:901] (1/2) Epoch 31, batch 2500, loss[loss=0.1371, simple_loss=0.2265, pruned_loss=0.02388, over 7241.00 frames. ], tot_loss[loss=0.1359, simple_loss=0.217, pruned_loss=0.02742, over 1442373.89 frames. ], batch size: 55, lr: 5.30e-03, grad_scale: 16.0 +2023-03-21 06:12:02,618 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87230.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:12:04,590 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87234.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:12:05,605 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87236.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:12:05,989 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.248e+02 1.655e+02 1.992e+02 2.464e+02 4.160e+02, threshold=3.984e+02, percent-clipped=0.0 +2023-03-21 06:12:09,052 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 06:12:14,638 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4794, 1.8299, 1.3225, 1.5849, 1.9342, 1.4794, 1.6882, 1.1320], + device='cuda:1'), covar=tensor([0.0204, 0.0133, 0.0267, 0.0201, 0.0121, 0.0259, 0.0144, 0.0219], + device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0030, 0.0030, 0.0031, 0.0030, 0.0028, 0.0032, 0.0041], + device='cuda:1'), out_proj_covar=tensor([3.7852e-05, 3.3939e-05, 3.4381e-05, 3.4809e-05, 3.3960e-05, 3.1997e-05, + 3.6240e-05, 4.6098e-05], device='cuda:1') +2023-03-21 06:12:23,124 INFO [train.py:901] (1/2) Epoch 31, batch 2550, loss[loss=0.1675, simple_loss=0.2419, pruned_loss=0.04656, over 7280.00 frames. ], tot_loss[loss=0.1358, simple_loss=0.2168, pruned_loss=0.02739, over 1442823.19 frames. ], batch size: 57, lr: 5.30e-03, grad_scale: 16.0 +2023-03-21 06:12:27,421 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.71 vs. limit=5.0 +2023-03-21 06:12:28,723 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7950, 4.2878, 4.1614, 4.7858, 4.6373, 4.6785, 4.1190, 4.3521], + device='cuda:1'), covar=tensor([0.0878, 0.2722, 0.2305, 0.1081, 0.0928, 0.1471, 0.0995, 0.1236], + device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0378, 0.0283, 0.0289, 0.0213, 0.0357, 0.0216, 0.0261], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:12:29,715 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=87284.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:12:33,916 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87291.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:12:49,290 INFO [train.py:901] (1/2) Epoch 31, batch 2600, loss[loss=0.1121, simple_loss=0.193, pruned_loss=0.01561, over 7201.00 frames. ], tot_loss[loss=0.1355, simple_loss=0.2163, pruned_loss=0.02738, over 1443699.83 frames. ], batch size: 39, lr: 5.29e-03, grad_scale: 16.0 +2023-03-21 06:12:57,248 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.274e+02 1.776e+02 2.100e+02 2.547e+02 3.881e+02, threshold=4.201e+02, percent-clipped=0.0 +2023-03-21 06:13:07,903 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87358.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:13:09,955 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87362.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:13:14,184 INFO [train.py:901] (1/2) Epoch 31, batch 2650, loss[loss=0.1467, simple_loss=0.2245, pruned_loss=0.03447, over 7323.00 frames. ], tot_loss[loss=0.136, simple_loss=0.2167, pruned_loss=0.02766, over 1444194.55 frames. ], batch size: 61, lr: 5.29e-03, grad_scale: 16.0 +2023-03-21 06:13:16,716 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7592, 2.9430, 3.6378, 3.7211, 3.7203, 3.7990, 3.8222, 3.6561], + device='cuda:1'), covar=tensor([0.0033, 0.0141, 0.0038, 0.0035, 0.0037, 0.0034, 0.0044, 0.0054], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0068, 0.0056, 0.0054, 0.0054, 0.0059, 0.0049, 0.0073], + device='cuda:1'), out_proj_covar=tensor([8.2578e-05, 1.4413e-04, 1.0684e-04, 9.7936e-05, 9.6200e-05, 1.0927e-04, + 9.9651e-05, 1.4134e-04], device='cuda:1') +2023-03-21 06:13:32,274 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5466, 2.8141, 2.4281, 2.7168, 2.6215, 2.2699, 2.7624, 2.5694], + device='cuda:1'), covar=tensor([0.0841, 0.0575, 0.0932, 0.0827, 0.1438, 0.0814, 0.0859, 0.0939], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0055, 0.0064, 0.0055, 0.0054, 0.0057, 0.0055, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:13:39,056 INFO [train.py:901] (1/2) Epoch 31, batch 2700, loss[loss=0.1435, simple_loss=0.2153, pruned_loss=0.03592, over 7216.00 frames. ], tot_loss[loss=0.136, simple_loss=0.2167, pruned_loss=0.0277, over 1443925.88 frames. ], batch size: 50, lr: 5.29e-03, grad_scale: 16.0 +2023-03-21 06:13:40,170 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87423.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:13:41,188 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0438, 2.8304, 2.9014, 2.9999, 2.7094, 2.5809, 3.1601, 2.3422], + device='cuda:1'), covar=tensor([0.0548, 0.0728, 0.0681, 0.0656, 0.0607, 0.0921, 0.0689, 0.1789], + device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0339, 0.0274, 0.0357, 0.0294, 0.0291, 0.0343, 0.0262], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:13:46,823 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 9.848e+01 1.823e+02 2.135e+02 2.461e+02 5.467e+02, threshold=4.269e+02, percent-clipped=3.0 +2023-03-21 06:13:56,823 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1679, 3.7170, 3.8881, 3.8579, 3.8250, 3.7012, 4.0793, 3.6001], + device='cuda:1'), covar=tensor([0.0121, 0.0219, 0.0105, 0.0165, 0.0432, 0.0120, 0.0136, 0.0185], + device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0099, 0.0097, 0.0085, 0.0171, 0.0105, 0.0103, 0.0108], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:13:58,775 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87461.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:14:03,325 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87470.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:14:03,722 INFO [train.py:901] (1/2) Epoch 31, batch 2750, loss[loss=0.1381, simple_loss=0.2245, pruned_loss=0.02582, over 7318.00 frames. ], tot_loss[loss=0.1365, simple_loss=0.2172, pruned_loss=0.02795, over 1444463.96 frames. ], batch size: 75, lr: 5.29e-03, grad_scale: 16.0 +2023-03-21 06:14:20,512 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87505.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:14:26,771 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=87518.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:14:28,552 INFO [train.py:901] (1/2) Epoch 31, batch 2800, loss[loss=0.1253, simple_loss=0.2009, pruned_loss=0.02491, over 7256.00 frames. ], tot_loss[loss=0.1361, simple_loss=0.2166, pruned_loss=0.02782, over 1441873.80 frames. ], batch size: 47, lr: 5.29e-03, grad_scale: 16.0 +2023-03-21 06:14:29,202 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87522.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:14:34,982 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87534.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:14:36,277 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.300e+02 1.877e+02 2.166e+02 2.662e+02 4.595e+02, threshold=4.333e+02, percent-clipped=1.0 +2023-03-21 06:14:54,744 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 06:14:54,801 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 06:14:54,828 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 06:14:54,845 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 06:14:55,020 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 06:14:55,037 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 06:14:55,047 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 06:14:55,149 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 06:14:55,225 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 06:14:55,241 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 06:14:55,258 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 06:14:55,727 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 06:14:55,733 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 06:14:55,740 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 06:14:55,881 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 06:14:55,884 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 06:14:55,941 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 06:15:00,752 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9290, 3.5186, 3.8557, 3.9730, 3.8992, 3.8620, 4.0160, 3.8034], + device='cuda:1'), covar=tensor([0.0029, 0.0086, 0.0033, 0.0028, 0.0029, 0.0033, 0.0036, 0.0046], + device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0066, 0.0055, 0.0054, 0.0053, 0.0059, 0.0048, 0.0072], + device='cuda:1'), out_proj_covar=tensor([8.1066e-05, 1.4148e-04, 1.0445e-04, 9.6330e-05, 9.4387e-05, 1.0817e-04, + 9.7628e-05, 1.3990e-04], device='cuda:1') +2023-03-21 06:15:01,126 INFO [train.py:901] (1/2) Epoch 32, batch 0, loss[loss=0.1316, simple_loss=0.2104, pruned_loss=0.02638, over 7321.00 frames. ], tot_loss[loss=0.1316, simple_loss=0.2104, pruned_loss=0.02638, over 7321.00 frames. ], batch size: 61, lr: 5.20e-03, grad_scale: 16.0 +2023-03-21 06:15:01,126 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 06:15:17,707 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1100, 3.7909, 3.8210, 3.8143, 3.8104, 3.7360, 3.9898, 3.4927], + device='cuda:1'), covar=tensor([0.0125, 0.0174, 0.0125, 0.0171, 0.0467, 0.0116, 0.0157, 0.0232], + device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0098, 0.0096, 0.0084, 0.0169, 0.0104, 0.0102, 0.0107], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:15:27,536 INFO [train.py:935] (1/2) Epoch 32, validation: loss=0.1652, simple_loss=0.2562, pruned_loss=0.03713, over 1622729.00 frames. +2023-03-21 06:15:27,537 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 06:15:30,231 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3697, 4.2672, 3.9831, 3.7876, 3.6097, 2.6746, 2.0709, 4.3635], + device='cuda:1'), covar=tensor([0.0040, 0.0067, 0.0071, 0.0070, 0.0109, 0.0442, 0.0559, 0.0044], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0087, 0.0106, 0.0090, 0.0121, 0.0129, 0.0126, 0.0099], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 06:15:30,805 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0159, 2.8656, 2.9883, 2.9306, 2.5969, 2.6391, 3.0847, 2.3527], + device='cuda:1'), covar=tensor([0.0494, 0.0542, 0.0566, 0.0649, 0.0598, 0.0881, 0.0647, 0.1736], + device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0334, 0.0271, 0.0354, 0.0290, 0.0288, 0.0340, 0.0257], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:15:33,114 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 06:15:35,683 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1040, 3.7727, 3.8260, 3.8289, 3.7029, 3.6943, 4.0255, 3.5272], + device='cuda:1'), covar=tensor([0.0161, 0.0183, 0.0129, 0.0172, 0.0484, 0.0135, 0.0147, 0.0220], + device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0099, 0.0096, 0.0085, 0.0171, 0.0105, 0.0103, 0.0108], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:15:38,167 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87566.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:15:43,590 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 06:15:46,179 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=87582.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:15:47,318 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87584.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:15:48,274 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87586.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:15:49,867 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87589.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:15:50,276 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 06:15:51,897 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3319, 3.3460, 3.3388, 3.2073, 3.3919, 3.2712, 2.7349, 3.4549], + device='cuda:1'), covar=tensor([0.1436, 0.0793, 0.1531, 0.2078, 0.1089, 0.1742, 0.3317, 0.1917], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0050, 0.0048, 0.0048, 0.0048, 0.0068, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 06:15:52,804 INFO [train.py:901] (1/2) Epoch 32, batch 50, loss[loss=0.1397, simple_loss=0.223, pruned_loss=0.02817, over 7247.00 frames. ], tot_loss[loss=0.1361, simple_loss=0.2182, pruned_loss=0.02698, over 325683.18 frames. ], batch size: 55, lr: 5.20e-03, grad_scale: 16.0 +2023-03-21 06:15:52,809 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 06:15:55,374 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 06:15:58,307 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87605.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:16:13,409 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 06:16:14,552 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 06:16:15,198 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4483, 2.3241, 2.2351, 3.5575, 1.6992, 3.2516, 1.4536, 3.1508], + device='cuda:1'), covar=tensor([0.0210, 0.1368, 0.1667, 0.0212, 0.3722, 0.0245, 0.1206, 0.0395], + device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0254, 0.0267, 0.0206, 0.0256, 0.0213, 0.0236, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 06:16:15,394 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 +2023-03-21 06:16:15,506 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.185e+02 1.723e+02 1.992e+02 2.354e+02 4.826e+02, threshold=3.984e+02, percent-clipped=1.0 +2023-03-21 06:16:19,616 INFO [train.py:901] (1/2) Epoch 32, batch 100, loss[loss=0.1345, simple_loss=0.2156, pruned_loss=0.02669, over 7294.00 frames. ], tot_loss[loss=0.135, simple_loss=0.2162, pruned_loss=0.02694, over 572876.27 frames. ], batch size: 80, lr: 5.20e-03, grad_scale: 16.0 +2023-03-21 06:16:19,754 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87645.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 06:16:22,316 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87650.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:16:26,102 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87658.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 06:16:30,198 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87666.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 06:16:40,963 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 +2023-03-21 06:16:44,669 INFO [train.py:901] (1/2) Epoch 32, batch 150, loss[loss=0.1394, simple_loss=0.2257, pruned_loss=0.02657, over 7341.00 frames. ], tot_loss[loss=0.1342, simple_loss=0.215, pruned_loss=0.02673, over 767387.90 frames. ], batch size: 61, lr: 5.20e-03, grad_scale: 16.0 +2023-03-21 06:16:50,353 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=87706.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:16:56,363 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87718.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:17:07,051 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.194e+02 1.862e+02 2.197e+02 2.502e+02 5.473e+02, threshold=4.393e+02, percent-clipped=2.0 +2023-03-21 06:17:11,197 INFO [train.py:901] (1/2) Epoch 32, batch 200, loss[loss=0.1345, simple_loss=0.2107, pruned_loss=0.02917, over 7204.00 frames. ], tot_loss[loss=0.1348, simple_loss=0.2152, pruned_loss=0.02717, over 916635.71 frames. ], batch size: 45, lr: 5.20e-03, grad_scale: 16.0 +2023-03-21 06:17:16,200 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 06:17:21,906 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 06:17:27,988 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 06:17:36,374 INFO [train.py:901] (1/2) Epoch 32, batch 250, loss[loss=0.1503, simple_loss=0.2346, pruned_loss=0.03299, over 7220.00 frames. ], tot_loss[loss=0.135, simple_loss=0.2153, pruned_loss=0.02735, over 1033869.60 frames. ], batch size: 93, lr: 5.20e-03, grad_scale: 16.0 +2023-03-21 06:17:40,636 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 06:17:40,709 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3236, 4.8369, 4.8769, 4.8609, 4.7455, 4.2653, 4.9082, 4.7237], + device='cuda:1'), covar=tensor([0.0470, 0.0350, 0.0394, 0.0457, 0.0319, 0.0440, 0.0323, 0.0449], + device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0248, 0.0190, 0.0195, 0.0150, 0.0223, 0.0196, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:17:48,957 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87817.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:17:54,717 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-03-21 06:17:58,941 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.241e+02 1.893e+02 2.142e+02 2.477e+02 4.691e+02, threshold=4.285e+02, percent-clipped=1.0 +2023-03-21 06:18:03,028 INFO [train.py:901] (1/2) Epoch 32, batch 300, loss[loss=0.1246, simple_loss=0.2061, pruned_loss=0.02156, over 7165.00 frames. ], tot_loss[loss=0.1351, simple_loss=0.2153, pruned_loss=0.02742, over 1124802.20 frames. ], batch size: 41, lr: 5.20e-03, grad_scale: 16.0 +2023-03-21 06:18:03,542 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 06:18:11,133 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 06:18:11,197 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87861.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 06:18:22,505 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 +2023-03-21 06:18:23,727 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87886.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:18:28,652 INFO [train.py:901] (1/2) Epoch 32, batch 350, loss[loss=0.1354, simple_loss=0.2133, pruned_loss=0.02872, over 7329.00 frames. ], tot_loss[loss=0.1352, simple_loss=0.2155, pruned_loss=0.02739, over 1196118.93 frames. ], batch size: 59, lr: 5.19e-03, grad_scale: 16.0 +2023-03-21 06:18:45,906 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 06:18:49,007 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=87934.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:18:50,420 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 1.703e+02 1.990e+02 2.314e+02 3.102e+02, threshold=3.979e+02, percent-clipped=0.0 +2023-03-21 06:18:52,144 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87940.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 06:18:54,637 INFO [train.py:901] (1/2) Epoch 32, batch 400, loss[loss=0.1438, simple_loss=0.23, pruned_loss=0.02885, over 7305.00 frames. ], tot_loss[loss=0.1358, simple_loss=0.2163, pruned_loss=0.02762, over 1252410.29 frames. ], batch size: 86, lr: 5.19e-03, grad_scale: 16.0 +2023-03-21 06:18:54,729 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87945.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:19:02,739 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87961.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:19:06,346 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3215, 2.6262, 2.8165, 2.5428, 2.9867, 2.5523, 1.8969, 2.3538], + device='cuda:1'), covar=tensor([0.0518, 0.0423, 0.0287, 0.0282, 0.0288, 0.0650, 0.0573, 0.0346], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0034, 0.0035, 0.0033, 0.0032, 0.0032, 0.0036, 0.0036], + device='cuda:1'), out_proj_covar=tensor([8.8675e-05, 8.8690e-05, 8.6992e-05, 8.3341e-05, 8.3737e-05, 8.4122e-05, + 8.9926e-05, 9.1910e-05], device='cuda:1') +2023-03-21 06:19:16,772 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2447, 2.8131, 3.2186, 3.0959, 3.3233, 2.9188, 2.7114, 3.2035], + device='cuda:1'), covar=tensor([0.1205, 0.0704, 0.1011, 0.1244, 0.0845, 0.1180, 0.2099, 0.1383], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0063, 0.0049, 0.0047, 0.0047, 0.0047, 0.0065, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 06:19:21,169 INFO [train.py:901] (1/2) Epoch 32, batch 450, loss[loss=0.1102, simple_loss=0.1869, pruned_loss=0.01674, over 6983.00 frames. ], tot_loss[loss=0.1358, simple_loss=0.2164, pruned_loss=0.02763, over 1292370.95 frames. ], batch size: 35, lr: 5.19e-03, grad_scale: 8.0 +2023-03-21 06:19:21,338 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7239, 1.4851, 1.9824, 2.2794, 2.0476, 2.2883, 1.9182, 2.2928], + device='cuda:1'), covar=tensor([0.2648, 0.3350, 0.3505, 0.2996, 0.2096, 0.1505, 0.2644, 0.1853], + device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0074, 0.0063, 0.0060, 0.0059, 0.0059, 0.0095, 0.0060], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:19:31,912 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 06:19:32,486 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 06:19:37,036 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88018.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:19:47,019 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.188e+02 1.770e+02 2.090e+02 2.498e+02 3.741e+02, threshold=4.181e+02, percent-clipped=0.0 +2023-03-21 06:19:50,539 INFO [train.py:901] (1/2) Epoch 32, batch 500, loss[loss=0.1355, simple_loss=0.2186, pruned_loss=0.02615, over 7229.00 frames. ], tot_loss[loss=0.1358, simple_loss=0.2164, pruned_loss=0.02765, over 1325575.42 frames. ], batch size: 93, lr: 5.19e-03, grad_scale: 8.0 +2023-03-21 06:19:57,398 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-03-21 06:20:00,173 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8042, 4.3140, 4.1864, 4.2757, 4.2997, 3.9390, 4.3604, 4.1959], + device='cuda:1'), covar=tensor([0.1187, 0.1097, 0.1149, 0.1111, 0.0838, 0.0801, 0.0822, 0.1035], + device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0248, 0.0190, 0.0195, 0.0150, 0.0222, 0.0196, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:20:01,225 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=88066.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:20:06,912 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 06:20:08,493 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 06:20:08,974 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 06:20:11,610 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 06:20:16,114 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 06:20:17,096 INFO [train.py:901] (1/2) Epoch 32, batch 550, loss[loss=0.1365, simple_loss=0.2201, pruned_loss=0.02641, over 7322.00 frames. ], tot_loss[loss=0.1361, simple_loss=0.2163, pruned_loss=0.02799, over 1348681.78 frames. ], batch size: 59, lr: 5.19e-03, grad_scale: 8.0 +2023-03-21 06:20:19,780 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88100.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:20:26,771 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 06:20:28,400 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88117.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:20:35,869 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 06:20:38,884 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 1.728e+02 2.085e+02 2.594e+02 4.584e+02, threshold=4.170e+02, percent-clipped=1.0 +2023-03-21 06:20:38,922 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 06:20:42,460 INFO [train.py:901] (1/2) Epoch 32, batch 600, loss[loss=0.139, simple_loss=0.2213, pruned_loss=0.02833, over 7321.00 frames. ], tot_loss[loss=0.1357, simple_loss=0.2161, pruned_loss=0.02765, over 1368377.55 frames. ], batch size: 59, lr: 5.19e-03, grad_scale: 8.0 +2023-03-21 06:20:45,377 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 06:20:51,708 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88161.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 06:20:51,736 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88161.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:20:53,659 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=88165.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:21:03,179 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 06:21:08,619 INFO [train.py:901] (1/2) Epoch 32, batch 650, loss[loss=0.1309, simple_loss=0.2129, pruned_loss=0.02444, over 7316.00 frames. ], tot_loss[loss=0.1358, simple_loss=0.2162, pruned_loss=0.0277, over 1384587.70 frames. ], batch size: 59, lr: 5.19e-03, grad_scale: 8.0 +2023-03-21 06:21:12,006 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 06:21:16,008 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=88209.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:21:25,573 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0064, 4.0118, 3.3421, 3.4524, 2.9556, 2.0906, 1.6945, 3.9637], + device='cuda:1'), covar=tensor([0.0049, 0.0047, 0.0122, 0.0071, 0.0175, 0.0558, 0.0671, 0.0053], + device='cuda:1'), in_proj_covar=tensor([0.0106, 0.0087, 0.0106, 0.0091, 0.0121, 0.0129, 0.0127, 0.0100], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 06:21:29,002 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 06:21:30,503 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.043e+02 1.734e+02 2.072e+02 2.519e+02 4.714e+02, threshold=4.144e+02, percent-clipped=1.0 +2023-03-21 06:21:31,623 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88240.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:21:34,618 INFO [train.py:901] (1/2) Epoch 32, batch 700, loss[loss=0.1414, simple_loss=0.221, pruned_loss=0.03091, over 7250.00 frames. ], tot_loss[loss=0.1354, simple_loss=0.2159, pruned_loss=0.02748, over 1398231.07 frames. ], batch size: 55, lr: 5.18e-03, grad_scale: 8.0 +2023-03-21 06:21:35,363 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88245.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:21:39,279 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 06:21:43,428 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88261.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:21:44,370 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0283, 3.8494, 3.7553, 3.8525, 3.1756, 3.6642, 3.9442, 3.5832], + device='cuda:1'), covar=tensor([0.0196, 0.0186, 0.0156, 0.0184, 0.0685, 0.0149, 0.0218, 0.0197], + device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0099, 0.0097, 0.0084, 0.0171, 0.0104, 0.0102, 0.0107], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:21:56,774 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=88288.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:21:58,404 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88291.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 06:21:59,321 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=88293.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:22:00,250 INFO [train.py:901] (1/2) Epoch 32, batch 750, loss[loss=0.1578, simple_loss=0.2464, pruned_loss=0.03455, over 6773.00 frames. ], tot_loss[loss=0.1357, simple_loss=0.2163, pruned_loss=0.02759, over 1409227.74 frames. ], batch size: 106, lr: 5.18e-03, grad_scale: 8.0 +2023-03-21 06:22:02,192 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 06:22:02,699 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 06:22:07,277 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=88309.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:22:15,843 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 06:22:21,098 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 06:22:23,015 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.235e+02 1.782e+02 2.097e+02 2.465e+02 5.037e+02, threshold=4.194e+02, percent-clipped=2.0 +2023-03-21 06:22:26,542 INFO [train.py:901] (1/2) Epoch 32, batch 800, loss[loss=0.1253, simple_loss=0.2036, pruned_loss=0.02345, over 7319.00 frames. ], tot_loss[loss=0.1359, simple_loss=0.2162, pruned_loss=0.02778, over 1415969.75 frames. ], batch size: 75, lr: 5.18e-03, grad_scale: 8.0 +2023-03-21 06:22:26,567 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 06:22:28,075 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 06:22:30,173 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88352.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:22:37,483 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 06:22:43,104 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4767, 4.0373, 3.9799, 4.4337, 4.3461, 4.4151, 3.9133, 3.9667], + device='cuda:1'), covar=tensor([0.0869, 0.2768, 0.2388, 0.1294, 0.0993, 0.1371, 0.0901, 0.1196], + device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0382, 0.0288, 0.0294, 0.0219, 0.0360, 0.0216, 0.0264], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:22:51,508 INFO [train.py:901] (1/2) Epoch 32, batch 850, loss[loss=0.1382, simple_loss=0.2201, pruned_loss=0.02817, over 7323.00 frames. ], tot_loss[loss=0.1351, simple_loss=0.2156, pruned_loss=0.02733, over 1419736.89 frames. ], batch size: 75, lr: 5.18e-03, grad_scale: 8.0 +2023-03-21 06:22:55,912 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 06:22:55,925 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 06:23:02,526 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 06:23:07,142 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 06:23:14,806 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.241e+02 1.715e+02 2.087e+02 2.362e+02 4.389e+02, threshold=4.174e+02, percent-clipped=1.0 +2023-03-21 06:23:18,456 INFO [train.py:901] (1/2) Epoch 32, batch 900, loss[loss=0.13, simple_loss=0.207, pruned_loss=0.02646, over 7299.00 frames. ], tot_loss[loss=0.1349, simple_loss=0.2152, pruned_loss=0.02729, over 1425767.09 frames. ], batch size: 49, lr: 5.18e-03, grad_scale: 8.0 +2023-03-21 06:23:24,125 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88456.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:23:40,548 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4616, 2.3570, 2.3287, 3.6839, 1.7777, 3.4000, 1.5695, 3.1797], + device='cuda:1'), covar=tensor([0.0184, 0.1307, 0.1680, 0.0169, 0.3646, 0.0201, 0.1094, 0.0412], + device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0251, 0.0263, 0.0201, 0.0251, 0.0210, 0.0234, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 06:23:43,379 INFO [train.py:901] (1/2) Epoch 32, batch 950, loss[loss=0.137, simple_loss=0.2233, pruned_loss=0.02532, over 7120.00 frames. ], tot_loss[loss=0.1345, simple_loss=0.2149, pruned_loss=0.02711, over 1430192.72 frames. ], batch size: 98, lr: 5.18e-03, grad_scale: 8.0 +2023-03-21 06:23:43,391 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 06:23:47,091 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 +2023-03-21 06:24:06,410 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.126e+02 1.732e+02 1.981e+02 2.527e+02 4.373e+02, threshold=3.962e+02, percent-clipped=2.0 +2023-03-21 06:24:08,399 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 06:24:09,903 INFO [train.py:901] (1/2) Epoch 32, batch 1000, loss[loss=0.1241, simple_loss=0.2092, pruned_loss=0.01953, over 7130.00 frames. ], tot_loss[loss=0.1348, simple_loss=0.2154, pruned_loss=0.02711, over 1432458.73 frames. ], batch size: 41, lr: 5.18e-03, grad_scale: 8.0 +2023-03-21 06:24:14,221 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-21 06:24:21,529 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2688, 4.1984, 3.8016, 3.7214, 3.3690, 2.5371, 2.1082, 4.2661], + device='cuda:1'), covar=tensor([0.0040, 0.0062, 0.0082, 0.0071, 0.0107, 0.0425, 0.0522, 0.0042], + device='cuda:1'), in_proj_covar=tensor([0.0106, 0.0086, 0.0106, 0.0091, 0.0120, 0.0129, 0.0126, 0.0099], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 06:24:28,541 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 06:24:36,349 INFO [train.py:901] (1/2) Epoch 32, batch 1050, loss[loss=0.1407, simple_loss=0.2278, pruned_loss=0.0268, over 7113.00 frames. ], tot_loss[loss=0.1351, simple_loss=0.216, pruned_loss=0.02707, over 1434267.72 frames. ], batch size: 98, lr: 5.17e-03, grad_scale: 8.0 +2023-03-21 06:24:50,691 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 06:24:54,355 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1069, 3.0216, 3.2141, 3.3443, 3.3045, 3.0939, 2.6682, 3.3254], + device='cuda:1'), covar=tensor([0.1523, 0.0677, 0.1325, 0.1103, 0.0940, 0.1215, 0.2321, 0.1195], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0064, 0.0050, 0.0048, 0.0048, 0.0048, 0.0067, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 06:24:54,763 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 06:24:58,204 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.073e+02 1.638e+02 1.910e+02 2.335e+02 5.404e+02, threshold=3.819e+02, percent-clipped=3.0 +2023-03-21 06:25:01,809 INFO [train.py:901] (1/2) Epoch 32, batch 1100, loss[loss=0.1166, simple_loss=0.2047, pruned_loss=0.01426, over 7129.00 frames. ], tot_loss[loss=0.1353, simple_loss=0.2162, pruned_loss=0.02719, over 1435484.08 frames. ], batch size: 41, lr: 5.17e-03, grad_scale: 8.0 +2023-03-21 06:25:02,934 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88647.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:25:05,045 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7660, 3.0583, 2.6852, 3.0095, 2.9888, 2.7364, 2.9701, 2.9048], + device='cuda:1'), covar=tensor([0.0807, 0.0503, 0.0746, 0.1273, 0.1307, 0.0670, 0.0839, 0.0758], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0056, 0.0064, 0.0056, 0.0053, 0.0057, 0.0055, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:25:08,571 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1612, 4.0955, 3.4715, 3.6588, 3.2536, 2.4050, 1.9035, 4.1722], + device='cuda:1'), covar=tensor([0.0046, 0.0057, 0.0113, 0.0065, 0.0122, 0.0449, 0.0597, 0.0045], + device='cuda:1'), in_proj_covar=tensor([0.0106, 0.0086, 0.0105, 0.0090, 0.0120, 0.0128, 0.0125, 0.0099], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 06:25:18,919 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3980, 1.6397, 1.5572, 1.5220, 1.6859, 1.4887, 1.4711, 1.1194], + device='cuda:1'), covar=tensor([0.0179, 0.0130, 0.0152, 0.0151, 0.0103, 0.0169, 0.0135, 0.0156], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0031, 0.0031, 0.0032, 0.0030, 0.0029, 0.0033, 0.0040], + device='cuda:1'), out_proj_covar=tensor([3.8330e-05, 3.5323e-05, 3.5360e-05, 3.5781e-05, 3.3995e-05, 3.2926e-05, + 3.7772e-05, 4.4987e-05], device='cuda:1') +2023-03-21 06:25:24,167 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 06:25:24,659 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 06:25:25,411 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 +2023-03-21 06:25:28,141 INFO [train.py:901] (1/2) Epoch 32, batch 1150, loss[loss=0.1436, simple_loss=0.2233, pruned_loss=0.03196, over 7271.00 frames. ], tot_loss[loss=0.1349, simple_loss=0.2156, pruned_loss=0.02708, over 1437459.71 frames. ], batch size: 57, lr: 5.17e-03, grad_scale: 8.0 +2023-03-21 06:25:36,086 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 06:25:36,584 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 06:25:49,710 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.246e+02 1.764e+02 2.030e+02 2.490e+02 8.067e+02, threshold=4.061e+02, percent-clipped=2.0 +2023-03-21 06:25:53,217 INFO [train.py:901] (1/2) Epoch 32, batch 1200, loss[loss=0.1321, simple_loss=0.2147, pruned_loss=0.02479, over 7309.00 frames. ], tot_loss[loss=0.1351, simple_loss=0.2157, pruned_loss=0.02722, over 1436700.76 frames. ], batch size: 49, lr: 5.17e-03, grad_scale: 8.0 +2023-03-21 06:25:55,295 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5002, 4.0265, 3.9552, 4.4982, 4.3239, 4.4395, 3.7962, 4.0753], + device='cuda:1'), covar=tensor([0.0818, 0.2615, 0.2134, 0.1158, 0.0992, 0.1308, 0.1034, 0.1275], + device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0385, 0.0290, 0.0293, 0.0219, 0.0363, 0.0216, 0.0266], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:25:58,185 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-21 06:26:00,108 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88756.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:26:03,538 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5693, 4.1150, 4.0276, 4.5598, 4.3971, 4.5103, 3.8383, 4.0843], + device='cuda:1'), covar=tensor([0.0791, 0.2536, 0.2204, 0.1050, 0.0940, 0.1221, 0.0889, 0.1147], + device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0386, 0.0291, 0.0294, 0.0219, 0.0364, 0.0217, 0.0266], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:26:09,531 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 06:26:19,486 INFO [train.py:901] (1/2) Epoch 32, batch 1250, loss[loss=0.1148, simple_loss=0.1873, pruned_loss=0.02115, over 6997.00 frames. ], tot_loss[loss=0.1351, simple_loss=0.2157, pruned_loss=0.0272, over 1438504.61 frames. ], batch size: 35, lr: 5.17e-03, grad_scale: 8.0 +2023-03-21 06:26:24,496 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=88804.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:26:33,453 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 06:26:37,448 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 06:26:38,455 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 06:26:41,997 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.302e+02 1.883e+02 2.188e+02 2.515e+02 4.579e+02, threshold=4.376e+02, percent-clipped=2.0 +2023-03-21 06:26:46,153 INFO [train.py:901] (1/2) Epoch 32, batch 1300, loss[loss=0.1371, simple_loss=0.2217, pruned_loss=0.02622, over 7234.00 frames. ], tot_loss[loss=0.1349, simple_loss=0.2153, pruned_loss=0.0272, over 1439110.86 frames. ], batch size: 55, lr: 5.17e-03, grad_scale: 8.0 +2023-03-21 06:27:01,563 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 06:27:03,581 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 06:27:06,608 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 06:27:11,021 INFO [train.py:901] (1/2) Epoch 32, batch 1350, loss[loss=0.1372, simple_loss=0.2229, pruned_loss=0.0258, over 7348.00 frames. ], tot_loss[loss=0.1356, simple_loss=0.2162, pruned_loss=0.02748, over 1438825.26 frames. ], batch size: 54, lr: 5.17e-03, grad_scale: 8.0 +2023-03-21 06:27:16,058 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 06:27:31,781 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1742, 3.7851, 3.7649, 3.7855, 3.2110, 3.6321, 4.1067, 3.5215], + device='cuda:1'), covar=tensor([0.0263, 0.0246, 0.0226, 0.0267, 0.0972, 0.0238, 0.0309, 0.0323], + device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0101, 0.0098, 0.0087, 0.0174, 0.0106, 0.0104, 0.0110], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:27:33,607 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.258e+02 1.816e+02 2.111e+02 2.445e+02 5.910e+02, threshold=4.221e+02, percent-clipped=1.0 +2023-03-21 06:27:37,100 INFO [train.py:901] (1/2) Epoch 32, batch 1400, loss[loss=0.1528, simple_loss=0.2457, pruned_loss=0.02991, over 7329.00 frames. ], tot_loss[loss=0.1352, simple_loss=0.2158, pruned_loss=0.02732, over 1441456.89 frames. ], batch size: 61, lr: 5.16e-03, grad_scale: 8.0 +2023-03-21 06:27:38,224 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88947.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:27:50,061 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 06:28:02,049 INFO [train.py:901] (1/2) Epoch 32, batch 1450, loss[loss=0.1443, simple_loss=0.2241, pruned_loss=0.03223, over 7238.00 frames. ], tot_loss[loss=0.1349, simple_loss=0.2156, pruned_loss=0.02712, over 1442649.09 frames. ], batch size: 45, lr: 5.16e-03, grad_scale: 8.0 +2023-03-21 06:28:02,108 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=88995.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:28:06,090 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-03-21 06:28:15,772 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9608, 2.4535, 1.9115, 2.8606, 2.6243, 2.4455, 2.5684, 2.6993], + device='cuda:1'), covar=tensor([0.2135, 0.1054, 0.3769, 0.0848, 0.0340, 0.0228, 0.0414, 0.0460], + device='cuda:1'), in_proj_covar=tensor([0.0259, 0.0237, 0.0256, 0.0263, 0.0193, 0.0191, 0.0213, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 06:28:16,114 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 06:28:19,808 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89027.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:28:20,847 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8697, 2.5630, 2.6007, 2.8387, 2.3628, 2.4476, 2.8211, 2.0623], + device='cuda:1'), covar=tensor([0.0643, 0.0737, 0.0693, 0.0779, 0.0700, 0.1016, 0.0900, 0.1984], + device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0336, 0.0272, 0.0358, 0.0290, 0.0291, 0.0343, 0.0258], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:28:25,190 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 1.675e+02 2.024e+02 2.357e+02 4.449e+02, threshold=4.048e+02, percent-clipped=2.0 +2023-03-21 06:28:28,795 INFO [train.py:901] (1/2) Epoch 32, batch 1500, loss[loss=0.1572, simple_loss=0.2396, pruned_loss=0.03742, over 6737.00 frames. ], tot_loss[loss=0.135, simple_loss=0.2155, pruned_loss=0.02725, over 1441207.74 frames. ], batch size: 107, lr: 5.16e-03, grad_scale: 8.0 +2023-03-21 06:28:28,951 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8595, 2.3115, 1.8572, 2.7286, 2.7290, 2.3489, 2.3791, 2.6770], + device='cuda:1'), covar=tensor([0.2148, 0.1077, 0.3687, 0.0731, 0.0287, 0.0215, 0.0389, 0.0404], + device='cuda:1'), in_proj_covar=tensor([0.0259, 0.0237, 0.0256, 0.0263, 0.0193, 0.0191, 0.0214, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 06:28:31,787 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 06:28:50,412 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89088.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:28:53,813 INFO [train.py:901] (1/2) Epoch 32, batch 1550, loss[loss=0.1469, simple_loss=0.2222, pruned_loss=0.03576, over 7330.00 frames. ], tot_loss[loss=0.1354, simple_loss=0.2161, pruned_loss=0.02737, over 1442858.88 frames. ], batch size: 54, lr: 5.16e-03, grad_scale: 8.0 +2023-03-21 06:28:55,537 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 06:29:06,723 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9311, 2.4133, 2.9920, 2.8358, 3.0780, 2.8003, 2.4490, 2.9886], + device='cuda:1'), covar=tensor([0.1198, 0.0756, 0.1208, 0.1523, 0.0730, 0.1097, 0.2089, 0.1253], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0063, 0.0049, 0.0047, 0.0047, 0.0047, 0.0066, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 06:29:16,715 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.184e+02 1.776e+02 2.055e+02 2.401e+02 3.292e+02, threshold=4.110e+02, percent-clipped=0.0 +2023-03-21 06:29:20,259 INFO [train.py:901] (1/2) Epoch 32, batch 1600, loss[loss=0.137, simple_loss=0.2217, pruned_loss=0.02618, over 7277.00 frames. ], tot_loss[loss=0.1357, simple_loss=0.2164, pruned_loss=0.02744, over 1443104.88 frames. ], batch size: 52, lr: 5.16e-03, grad_scale: 8.0 +2023-03-21 06:29:22,548 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 +2023-03-21 06:29:23,469 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9955, 3.6104, 3.4412, 3.6946, 3.2072, 3.1149, 3.9979, 2.7175], + device='cuda:1'), covar=tensor([0.0452, 0.0682, 0.0673, 0.0699, 0.0808, 0.1110, 0.0626, 0.2215], + device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0335, 0.0271, 0.0356, 0.0289, 0.0290, 0.0341, 0.0259], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:29:26,299 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 06:29:27,381 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 06:29:30,324 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 +2023-03-21 06:29:30,450 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 06:29:32,590 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89169.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:29:39,740 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 06:29:44,339 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 06:29:46,777 INFO [train.py:901] (1/2) Epoch 32, batch 1650, loss[loss=0.1485, simple_loss=0.2286, pruned_loss=0.0342, over 7263.00 frames. ], tot_loss[loss=0.136, simple_loss=0.2169, pruned_loss=0.02754, over 1443495.12 frames. ], batch size: 64, lr: 5.16e-03, grad_scale: 8.0 +2023-03-21 06:29:53,090 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 06:29:55,803 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.2185, 5.7313, 5.7499, 5.6762, 5.4598, 5.3383, 5.7999, 5.5689], + device='cuda:1'), covar=tensor([0.0343, 0.0254, 0.0256, 0.0331, 0.0247, 0.0234, 0.0197, 0.0340], + device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0247, 0.0187, 0.0191, 0.0149, 0.0221, 0.0193, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:30:04,908 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89230.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:30:08,767 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.241e+02 1.670e+02 2.019e+02 2.409e+02 3.267e+02, threshold=4.038e+02, percent-clipped=0.0 +2023-03-21 06:30:09,800 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 06:30:12,218 INFO [train.py:901] (1/2) Epoch 32, batch 1700, loss[loss=0.1351, simple_loss=0.2147, pruned_loss=0.02777, over 7224.00 frames. ], tot_loss[loss=0.1355, simple_loss=0.2163, pruned_loss=0.02736, over 1439362.19 frames. ], batch size: 45, lr: 5.15e-03, grad_scale: 8.0 +2023-03-21 06:30:13,698 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 06:30:22,536 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-21 06:30:24,756 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 06:30:24,926 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2406, 2.9028, 2.0491, 3.3347, 3.2580, 2.9569, 2.5369, 2.9136], + device='cuda:1'), covar=tensor([0.2082, 0.0882, 0.3680, 0.0582, 0.0298, 0.0223, 0.0457, 0.0521], + device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0237, 0.0253, 0.0262, 0.0192, 0.0190, 0.0213, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 06:30:38,376 INFO [train.py:901] (1/2) Epoch 32, batch 1750, loss[loss=0.1525, simple_loss=0.2368, pruned_loss=0.03412, over 7271.00 frames. ], tot_loss[loss=0.1362, simple_loss=0.2168, pruned_loss=0.02779, over 1440959.49 frames. ], batch size: 66, lr: 5.15e-03, grad_scale: 8.0 +2023-03-21 06:30:48,960 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7407, 3.1147, 2.6304, 2.8520, 2.9214, 2.6595, 2.9216, 2.8424], + device='cuda:1'), covar=tensor([0.1027, 0.0892, 0.0875, 0.1409, 0.1076, 0.0906, 0.0692, 0.1067], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0055, 0.0064, 0.0055, 0.0053, 0.0057, 0.0054, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:30:50,361 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 06:30:51,376 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 06:30:59,849 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.271e+02 1.754e+02 1.965e+02 2.362e+02 4.185e+02, threshold=3.929e+02, percent-clipped=1.0 +2023-03-21 06:31:02,017 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5416, 4.1009, 3.9653, 4.4595, 4.3265, 4.4592, 3.8727, 4.0166], + device='cuda:1'), covar=tensor([0.0803, 0.2434, 0.2212, 0.1150, 0.0879, 0.1185, 0.0959, 0.1345], + device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0381, 0.0288, 0.0295, 0.0219, 0.0359, 0.0215, 0.0266], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:31:03,411 INFO [train.py:901] (1/2) Epoch 32, batch 1800, loss[loss=0.1333, simple_loss=0.2176, pruned_loss=0.02448, over 7323.00 frames. ], tot_loss[loss=0.1358, simple_loss=0.2166, pruned_loss=0.02748, over 1442232.77 frames. ], batch size: 59, lr: 5.15e-03, grad_scale: 8.0 +2023-03-21 06:31:11,923 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 06:31:19,170 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89374.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:31:23,755 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89383.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:31:25,244 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 06:31:29,745 INFO [train.py:901] (1/2) Epoch 32, batch 1850, loss[loss=0.1221, simple_loss=0.1986, pruned_loss=0.02277, over 7183.00 frames. ], tot_loss[loss=0.1356, simple_loss=0.2163, pruned_loss=0.02743, over 1441263.99 frames. ], batch size: 39, lr: 5.15e-03, grad_scale: 8.0 +2023-03-21 06:31:35,125 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 06:31:45,161 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89425.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:31:48,819 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 +2023-03-21 06:31:50,120 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89435.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:31:51,495 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.222e+02 1.843e+02 2.192e+02 2.573e+02 5.735e+02, threshold=4.384e+02, percent-clipped=2.0 +2023-03-21 06:31:51,526 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 06:31:52,713 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89439.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:31:55,572 INFO [train.py:901] (1/2) Epoch 32, batch 1900, loss[loss=0.1369, simple_loss=0.2147, pruned_loss=0.02949, over 7368.00 frames. ], tot_loss[loss=0.1353, simple_loss=0.2161, pruned_loss=0.02728, over 1442428.17 frames. ], batch size: 51, lr: 5.15e-03, grad_scale: 8.0 +2023-03-21 06:32:00,790 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2151, 2.8758, 3.4001, 3.0987, 3.2530, 3.1937, 2.8379, 3.1704], + device='cuda:1'), covar=tensor([0.1906, 0.0813, 0.1060, 0.1397, 0.1094, 0.0905, 0.1505, 0.2043], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0065, 0.0050, 0.0048, 0.0048, 0.0048, 0.0066, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 06:32:13,596 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8081, 2.6311, 2.8200, 2.9576, 2.5963, 2.6064, 2.9279, 2.2724], + device='cuda:1'), covar=tensor([0.0536, 0.0561, 0.0611, 0.0658, 0.0585, 0.0791, 0.0573, 0.1706], + device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0335, 0.0272, 0.0358, 0.0290, 0.0292, 0.0345, 0.0258], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:32:17,050 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89486.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:32:17,877 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 06:32:21,390 INFO [train.py:901] (1/2) Epoch 32, batch 1950, loss[loss=0.1209, simple_loss=0.205, pruned_loss=0.01845, over 7315.00 frames. ], tot_loss[loss=0.1356, simple_loss=0.2163, pruned_loss=0.0275, over 1443620.23 frames. ], batch size: 59, lr: 5.15e-03, grad_scale: 8.0 +2023-03-21 06:32:24,033 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89500.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 06:32:28,459 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89509.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:32:28,820 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 06:32:33,298 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 06:32:33,794 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 06:32:36,985 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89525.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:32:44,150 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.288e+02 1.741e+02 2.051e+02 2.439e+02 5.519e+02, threshold=4.102e+02, percent-clipped=1.0 +2023-03-21 06:32:47,693 INFO [train.py:901] (1/2) Epoch 32, batch 2000, loss[loss=0.1288, simple_loss=0.207, pruned_loss=0.02532, over 7320.00 frames. ], tot_loss[loss=0.1362, simple_loss=0.2169, pruned_loss=0.02774, over 1444694.69 frames. ], batch size: 44, lr: 5.15e-03, grad_scale: 8.0 +2023-03-21 06:32:51,670 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 06:32:58,237 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0720, 4.0142, 3.4800, 3.6412, 3.0904, 2.3794, 1.7625, 4.0848], + device='cuda:1'), covar=tensor([0.0048, 0.0060, 0.0114, 0.0059, 0.0149, 0.0503, 0.0639, 0.0052], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0087, 0.0107, 0.0090, 0.0122, 0.0130, 0.0127, 0.0100], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 06:33:00,359 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89570.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:33:02,711 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 06:33:09,632 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 06:33:12,624 INFO [train.py:901] (1/2) Epoch 32, batch 2050, loss[loss=0.124, simple_loss=0.2071, pruned_loss=0.02042, over 7343.00 frames. ], tot_loss[loss=0.1365, simple_loss=0.2175, pruned_loss=0.02777, over 1444930.91 frames. ], batch size: 73, lr: 5.14e-03, grad_scale: 8.0 +2023-03-21 06:33:18,896 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 +2023-03-21 06:33:25,071 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 +2023-03-21 06:33:35,813 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.785e+02 2.119e+02 2.528e+02 6.748e+02, threshold=4.239e+02, percent-clipped=3.0 +2023-03-21 06:33:39,338 INFO [train.py:901] (1/2) Epoch 32, batch 2100, loss[loss=0.1233, simple_loss=0.2069, pruned_loss=0.01986, over 7269.00 frames. ], tot_loss[loss=0.1359, simple_loss=0.2168, pruned_loss=0.02751, over 1446527.79 frames. ], batch size: 64, lr: 5.14e-03, grad_scale: 8.0 +2023-03-21 06:33:45,877 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 06:33:48,371 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 06:33:48,505 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7466, 2.4851, 2.2304, 3.9102, 1.8717, 3.5896, 1.4429, 3.1758], + device='cuda:1'), covar=tensor([0.0179, 0.1228, 0.1825, 0.0174, 0.3772, 0.0217, 0.1265, 0.0309], + device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0249, 0.0263, 0.0202, 0.0252, 0.0210, 0.0232, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 06:33:57,154 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89680.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:33:58,671 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89683.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:34:02,379 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-03-21 06:34:04,476 INFO [train.py:901] (1/2) Epoch 32, batch 2150, loss[loss=0.1531, simple_loss=0.2341, pruned_loss=0.03603, over 7295.00 frames. ], tot_loss[loss=0.1359, simple_loss=0.2167, pruned_loss=0.02755, over 1448151.14 frames. ], batch size: 80, lr: 5.14e-03, grad_scale: 8.0 +2023-03-21 06:34:23,000 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 +2023-03-21 06:34:23,157 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89730.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:34:23,641 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=89731.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:34:27,156 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.261e+02 1.892e+02 2.197e+02 2.578e+02 4.314e+02, threshold=4.394e+02, percent-clipped=1.0 +2023-03-21 06:34:28,834 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89741.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:34:30,662 INFO [train.py:901] (1/2) Epoch 32, batch 2200, loss[loss=0.1296, simple_loss=0.2139, pruned_loss=0.02271, over 7317.00 frames. ], tot_loss[loss=0.1358, simple_loss=0.2165, pruned_loss=0.0276, over 1447412.65 frames. ], batch size: 75, lr: 5.14e-03, grad_scale: 8.0 +2023-03-21 06:34:33,540 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 06:34:42,852 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 +2023-03-21 06:34:48,506 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89781.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:34:53,684 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0168, 2.6240, 1.8248, 2.9712, 2.8331, 2.6732, 2.1825, 2.4865], + device='cuda:1'), covar=tensor([0.1960, 0.0978, 0.3792, 0.0586, 0.0251, 0.0209, 0.0247, 0.0353], + device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0236, 0.0252, 0.0260, 0.0191, 0.0190, 0.0212, 0.0222], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 06:34:55,984 INFO [train.py:901] (1/2) Epoch 32, batch 2250, loss[loss=0.1536, simple_loss=0.2351, pruned_loss=0.03604, over 7279.00 frames. ], tot_loss[loss=0.1362, simple_loss=0.2171, pruned_loss=0.02764, over 1448078.71 frames. ], batch size: 57, lr: 5.14e-03, grad_scale: 8.0 +2023-03-21 06:34:56,057 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89795.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 06:35:00,803 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 +2023-03-21 06:35:07,502 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 06:35:07,977 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 06:35:12,140 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89825.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:35:16,847 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.07 vs. limit=5.0 +2023-03-21 06:35:18,444 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.217e+02 1.744e+02 2.009e+02 2.399e+02 4.043e+02, threshold=4.017e+02, percent-clipped=0.0 +2023-03-21 06:35:19,970 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 06:35:21,897 INFO [train.py:901] (1/2) Epoch 32, batch 2300, loss[loss=0.1254, simple_loss=0.2087, pruned_loss=0.02103, over 7329.00 frames. ], tot_loss[loss=0.1364, simple_loss=0.2173, pruned_loss=0.02777, over 1448735.11 frames. ], batch size: 61, lr: 5.14e-03, grad_scale: 8.0 +2023-03-21 06:35:28,558 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8886, 2.1125, 2.1099, 2.0449, 2.0219, 2.0760, 1.7621, 1.8075], + device='cuda:1'), covar=tensor([0.0908, 0.0587, 0.0592, 0.0480, 0.0532, 0.0651, 0.0587, 0.0356], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0034, 0.0035, 0.0034, 0.0033, 0.0032, 0.0037, 0.0036], + device='cuda:1'), out_proj_covar=tensor([8.9783e-05, 8.8437e-05, 8.7238e-05, 8.5377e-05, 8.6129e-05, 8.3366e-05, + 9.1171e-05, 9.0790e-05], device='cuda:1') +2023-03-21 06:35:31,980 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:35:32,054 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:35:32,788 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-03-21 06:35:36,757 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=89873.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:35:48,390 INFO [train.py:901] (1/2) Epoch 32, batch 2350, loss[loss=0.1277, simple_loss=0.213, pruned_loss=0.02117, over 7267.00 frames. ], tot_loss[loss=0.136, simple_loss=0.217, pruned_loss=0.02747, over 1445964.20 frames. ], batch size: 52, lr: 5.14e-03, grad_scale: 8.0 +2023-03-21 06:35:56,661 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-03-21 06:36:04,013 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89926.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 06:36:06,407 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 06:36:09,752 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.391e+02 1.845e+02 2.244e+02 2.605e+02 4.330e+02, threshold=4.488e+02, percent-clipped=2.0 +2023-03-21 06:36:12,808 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 06:36:13,307 INFO [train.py:901] (1/2) Epoch 32, batch 2400, loss[loss=0.1354, simple_loss=0.2193, pruned_loss=0.02575, over 7339.00 frames. ], tot_loss[loss=0.1359, simple_loss=0.2171, pruned_loss=0.02736, over 1444459.51 frames. ], batch size: 61, lr: 5.13e-03, grad_scale: 8.0 +2023-03-21 06:36:23,495 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 06:36:25,309 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 +2023-03-21 06:36:26,440 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 06:36:39,643 INFO [train.py:901] (1/2) Epoch 32, batch 2450, loss[loss=0.1388, simple_loss=0.2243, pruned_loss=0.02666, over 7328.00 frames. ], tot_loss[loss=0.1359, simple_loss=0.2172, pruned_loss=0.02735, over 1444259.31 frames. ], batch size: 54, lr: 5.13e-03, grad_scale: 16.0 +2023-03-21 06:36:45,100 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4878, 3.9039, 4.1439, 4.2003, 4.1694, 4.0604, 4.4293, 3.8604], + device='cuda:1'), covar=tensor([0.0141, 0.0179, 0.0114, 0.0145, 0.0416, 0.0128, 0.0145, 0.0188], + device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0100, 0.0098, 0.0086, 0.0173, 0.0105, 0.0104, 0.0108], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:36:46,093 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7007, 3.0540, 3.5564, 3.7068, 3.7351, 3.7259, 3.6975, 3.6277], + device='cuda:1'), covar=tensor([0.0031, 0.0107, 0.0036, 0.0031, 0.0033, 0.0032, 0.0042, 0.0056], + device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0066, 0.0054, 0.0053, 0.0053, 0.0058, 0.0047, 0.0072], + device='cuda:1'), out_proj_covar=tensor([8.0006e-05, 1.3977e-04, 1.0277e-04, 9.4895e-05, 9.4655e-05, 1.0552e-04, + 9.4531e-05, 1.3838e-04], device='cuda:1') +2023-03-21 06:36:53,442 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 06:36:57,608 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90030.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:37:00,543 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90036.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:37:01,469 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.713e+02 2.021e+02 2.446e+02 5.091e+02, threshold=4.041e+02, percent-clipped=1.0 +2023-03-21 06:37:05,590 INFO [train.py:901] (1/2) Epoch 32, batch 2500, loss[loss=0.1092, simple_loss=0.1752, pruned_loss=0.02158, over 5905.00 frames. ], tot_loss[loss=0.1357, simple_loss=0.2167, pruned_loss=0.02741, over 1442934.11 frames. ], batch size: 25, lr: 5.13e-03, grad_scale: 16.0 +2023-03-21 06:37:18,230 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3241, 4.2408, 3.9573, 3.8226, 3.4504, 2.5959, 2.0815, 4.3819], + device='cuda:1'), covar=tensor([0.0046, 0.0062, 0.0095, 0.0070, 0.0114, 0.0494, 0.0575, 0.0043], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0086, 0.0106, 0.0090, 0.0121, 0.0129, 0.0125, 0.0099], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 06:37:18,393 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-03-21 06:37:20,142 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 06:37:22,760 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=90078.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:37:24,325 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90081.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:37:31,179 INFO [train.py:901] (1/2) Epoch 32, batch 2550, loss[loss=0.1289, simple_loss=0.2137, pruned_loss=0.02202, over 7297.00 frames. ], tot_loss[loss=0.1363, simple_loss=0.2172, pruned_loss=0.0277, over 1441546.49 frames. ], batch size: 86, lr: 5.13e-03, grad_scale: 16.0 +2023-03-21 06:37:31,293 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90095.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:37:46,285 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3966, 1.6470, 1.4572, 1.5639, 1.6458, 1.5428, 1.3775, 1.0379], + device='cuda:1'), covar=tensor([0.0154, 0.0141, 0.0260, 0.0143, 0.0100, 0.0120, 0.0138, 0.0192], + device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0031, 0.0031, 0.0032, 0.0030, 0.0029, 0.0033, 0.0041], + device='cuda:1'), out_proj_covar=tensor([3.8904e-05, 3.5207e-05, 3.5229e-05, 3.5998e-05, 3.4137e-05, 3.2897e-05, + 3.7858e-05, 4.5193e-05], device='cuda:1') +2023-03-21 06:37:48,871 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=90129.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:37:49,477 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90130.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:37:53,269 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.307e+02 1.845e+02 2.164e+02 2.494e+02 4.541e+02, threshold=4.327e+02, percent-clipped=1.0 +2023-03-21 06:37:55,814 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=90143.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 06:37:56,717 INFO [train.py:901] (1/2) Epoch 32, batch 2600, loss[loss=0.1308, simple_loss=0.2104, pruned_loss=0.02557, over 7292.00 frames. ], tot_loss[loss=0.1358, simple_loss=0.2166, pruned_loss=0.02745, over 1441559.50 frames. ], batch size: 66, lr: 5.13e-03, grad_scale: 16.0 +2023-03-21 06:38:06,770 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90165.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:38:18,142 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9576, 2.7896, 3.2172, 3.0700, 3.1424, 2.9235, 2.5654, 3.2671], + device='cuda:1'), covar=tensor([0.2152, 0.0687, 0.1320, 0.1341, 0.0925, 0.0980, 0.2238, 0.1158], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0063, 0.0049, 0.0047, 0.0047, 0.0047, 0.0064, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 06:38:20,095 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90191.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:38:20,225 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-03-21 06:38:21,899 INFO [train.py:901] (1/2) Epoch 32, batch 2650, loss[loss=0.1177, simple_loss=0.2027, pruned_loss=0.01634, over 7174.00 frames. ], tot_loss[loss=0.1359, simple_loss=0.217, pruned_loss=0.02744, over 1441519.18 frames. ], batch size: 39, lr: 5.13e-03, grad_scale: 16.0 +2023-03-21 06:38:31,049 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=90213.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:38:34,920 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90221.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 06:38:40,350 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90232.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:38:43,103 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.241e+02 1.716e+02 1.987e+02 2.459e+02 4.305e+02, threshold=3.973e+02, percent-clipped=0.0 +2023-03-21 06:38:43,893 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-03-21 06:38:46,626 INFO [train.py:901] (1/2) Epoch 32, batch 2700, loss[loss=0.1335, simple_loss=0.2141, pruned_loss=0.02648, over 7276.00 frames. ], tot_loss[loss=0.1361, simple_loss=0.2169, pruned_loss=0.02764, over 1440606.67 frames. ], batch size: 57, lr: 5.13e-03, grad_scale: 16.0 +2023-03-21 06:39:10,337 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90293.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:39:11,149 INFO [train.py:901] (1/2) Epoch 32, batch 2750, loss[loss=0.1449, simple_loss=0.219, pruned_loss=0.0354, over 7214.00 frames. ], tot_loss[loss=0.136, simple_loss=0.2167, pruned_loss=0.02768, over 1440819.21 frames. ], batch size: 45, lr: 5.12e-03, grad_scale: 16.0 +2023-03-21 06:39:16,947 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 +2023-03-21 06:39:31,490 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90336.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:39:32,299 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.279e+02 1.708e+02 1.999e+02 2.519e+02 4.921e+02, threshold=3.998e+02, percent-clipped=2.0 +2023-03-21 06:39:33,423 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4532, 2.5945, 3.3556, 3.4638, 3.4487, 3.4929, 3.2150, 3.3933], + device='cuda:1'), covar=tensor([0.0028, 0.0142, 0.0037, 0.0032, 0.0032, 0.0030, 0.0084, 0.0052], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0067, 0.0055, 0.0054, 0.0053, 0.0059, 0.0048, 0.0073], + device='cuda:1'), out_proj_covar=tensor([8.1427e-05, 1.4124e-04, 1.0308e-04, 9.6217e-05, 9.5346e-05, 1.0662e-04, + 9.6061e-05, 1.4001e-04], device='cuda:1') +2023-03-21 06:39:35,816 INFO [train.py:901] (1/2) Epoch 32, batch 2800, loss[loss=0.1458, simple_loss=0.2292, pruned_loss=0.03119, over 7239.00 frames. ], tot_loss[loss=0.1362, simple_loss=0.2166, pruned_loss=0.02791, over 1438458.42 frames. ], batch size: 55, lr: 5.12e-03, grad_scale: 16.0 +2023-03-21 06:39:38,831 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90351.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:40:00,531 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 06:40:00,890 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 06:40:00,919 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 06:40:00,935 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 06:40:01,103 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 06:40:01,118 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 06:40:01,128 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 06:40:01,233 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 06:40:01,307 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 06:40:01,323 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 06:40:01,338 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 06:40:01,509 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 06:40:01,514 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 06:40:01,521 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 06:40:02,011 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 06:40:02,013 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 06:40:02,070 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 06:40:06,627 INFO [train.py:901] (1/2) Epoch 33, batch 0, loss[loss=0.1029, simple_loss=0.1662, pruned_loss=0.01977, over 6343.00 frames. ], tot_loss[loss=0.1029, simple_loss=0.1662, pruned_loss=0.01977, over 6343.00 frames. ], batch size: 27, lr: 5.05e-03, grad_scale: 16.0 +2023-03-21 06:40:06,627 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 06:40:19,808 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6372, 2.8299, 2.5518, 2.8857, 2.8472, 2.5760, 2.8846, 2.6906], + device='cuda:1'), covar=tensor([0.0732, 0.0615, 0.1184, 0.0715, 0.1201, 0.0672, 0.0567, 0.0768], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0056, 0.0064, 0.0055, 0.0053, 0.0057, 0.0054, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:40:31,809 INFO [train.py:935] (1/2) Epoch 33, validation: loss=0.165, simple_loss=0.257, pruned_loss=0.03655, over 1622729.00 frames. +2023-03-21 06:40:31,810 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 06:40:38,931 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 06:40:39,967 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=90384.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:40:41,505 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9090, 3.9210, 3.3189, 3.5003, 2.8213, 2.2221, 1.7613, 3.9013], + device='cuda:1'), covar=tensor([0.0047, 0.0042, 0.0119, 0.0062, 0.0157, 0.0524, 0.0619, 0.0054], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0086, 0.0106, 0.0091, 0.0121, 0.0129, 0.0126, 0.0100], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 06:40:41,722 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-03-21 06:40:49,247 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 06:40:54,975 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90412.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:40:56,345 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 06:40:58,407 INFO [train.py:901] (1/2) Epoch 33, batch 50, loss[loss=0.1415, simple_loss=0.2249, pruned_loss=0.02901, over 7296.00 frames. ], tot_loss[loss=0.1355, simple_loss=0.2162, pruned_loss=0.02737, over 326054.86 frames. ], batch size: 86, lr: 5.04e-03, grad_scale: 16.0 +2023-03-21 06:40:58,428 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 06:41:00,918 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 06:41:07,836 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.447e+02 1.940e+02 2.295e+02 2.780e+02 4.742e+02, threshold=4.589e+02, percent-clipped=7.0 +2023-03-21 06:41:09,221 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.47 vs. limit=5.0 +2023-03-21 06:41:11,179 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-03-21 06:41:14,126 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0938, 2.8748, 3.2202, 3.0584, 2.8301, 2.8009, 3.1389, 2.4752], + device='cuda:1'), covar=tensor([0.0465, 0.0458, 0.0608, 0.0596, 0.0588, 0.0813, 0.0537, 0.1811], + device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0339, 0.0274, 0.0359, 0.0293, 0.0294, 0.0347, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:41:18,009 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 06:41:18,483 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 06:41:19,131 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90459.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:41:19,662 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3353, 2.8263, 2.0437, 3.2926, 3.2903, 3.1561, 2.8353, 2.7529], + device='cuda:1'), covar=tensor([0.1679, 0.0905, 0.3409, 0.0564, 0.0236, 0.0225, 0.0309, 0.0369], + device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0236, 0.0253, 0.0258, 0.0191, 0.0191, 0.0211, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 06:41:21,589 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3819, 3.9478, 4.0420, 4.0474, 3.9300, 3.9205, 4.2345, 3.7875], + device='cuda:1'), covar=tensor([0.0126, 0.0162, 0.0112, 0.0139, 0.0446, 0.0128, 0.0146, 0.0164], + device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0098, 0.0097, 0.0086, 0.0171, 0.0104, 0.0103, 0.0108], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:41:24,004 INFO [train.py:901] (1/2) Epoch 33, batch 100, loss[loss=0.1369, simple_loss=0.213, pruned_loss=0.03034, over 7261.00 frames. ], tot_loss[loss=0.1352, simple_loss=0.2166, pruned_loss=0.02695, over 574720.46 frames. ], batch size: 47, lr: 5.04e-03, grad_scale: 16.0 +2023-03-21 06:41:25,127 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7074, 3.0369, 2.6247, 2.9578, 2.8678, 2.5738, 2.9276, 2.6596], + device='cuda:1'), covar=tensor([0.0847, 0.0542, 0.0874, 0.0827, 0.1315, 0.0553, 0.0705, 0.1137], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0056, 0.0064, 0.0055, 0.0053, 0.0057, 0.0055, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:41:30,719 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.74 vs. limit=5.0 +2023-03-21 06:41:32,433 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90486.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:41:38,190 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90496.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:41:41,196 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3367, 2.5638, 2.2485, 2.4884, 2.5704, 2.1535, 2.6096, 2.5117], + device='cuda:1'), covar=tensor([0.0755, 0.0580, 0.1215, 0.0626, 0.1019, 0.1268, 0.0355, 0.0694], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0057, 0.0064, 0.0056, 0.0054, 0.0058, 0.0055, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:41:49,631 INFO [train.py:901] (1/2) Epoch 33, batch 150, loss[loss=0.1357, simple_loss=0.2219, pruned_loss=0.02473, over 7311.00 frames. ], tot_loss[loss=0.1355, simple_loss=0.2168, pruned_loss=0.02713, over 767616.61 frames. ], batch size: 75, lr: 5.04e-03, grad_scale: 16.0 +2023-03-21 06:41:50,313 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90520.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:41:50,771 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90521.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 06:41:59,746 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.776e+02 2.057e+02 2.515e+02 4.094e+02, threshold=4.114e+02, percent-clipped=0.0 +2023-03-21 06:42:09,418 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90557.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:42:15,303 INFO [train.py:901] (1/2) Epoch 33, batch 200, loss[loss=0.1234, simple_loss=0.2108, pruned_loss=0.018, over 7363.00 frames. ], tot_loss[loss=0.1357, simple_loss=0.2164, pruned_loss=0.02749, over 915932.36 frames. ], batch size: 73, lr: 5.04e-03, grad_scale: 16.0 +2023-03-21 06:42:15,371 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=90569.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:42:18,230 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 06:42:23,206 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 06:42:25,384 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90588.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:42:29,298 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 06:42:41,044 INFO [train.py:901] (1/2) Epoch 33, batch 250, loss[loss=0.1264, simple_loss=0.1992, pruned_loss=0.02679, over 7180.00 frames. ], tot_loss[loss=0.1352, simple_loss=0.2159, pruned_loss=0.02724, over 1033964.12 frames. ], batch size: 39, lr: 5.04e-03, grad_scale: 16.0 +2023-03-21 06:42:42,110 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 06:42:50,981 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.298e+02 1.739e+02 2.002e+02 2.238e+02 3.675e+02, threshold=4.004e+02, percent-clipped=0.0 +2023-03-21 06:43:01,601 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 06:43:06,556 INFO [train.py:901] (1/2) Epoch 33, batch 300, loss[loss=0.1423, simple_loss=0.2233, pruned_loss=0.0307, over 7220.00 frames. ], tot_loss[loss=0.1351, simple_loss=0.2161, pruned_loss=0.0271, over 1125627.95 frames. ], batch size: 50, lr: 5.04e-03, grad_scale: 8.0 +2023-03-21 06:43:10,679 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 06:43:27,020 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90707.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:43:32,916 INFO [train.py:901] (1/2) Epoch 33, batch 350, loss[loss=0.1272, simple_loss=0.2046, pruned_loss=0.02491, over 7265.00 frames. ], tot_loss[loss=0.1353, simple_loss=0.2159, pruned_loss=0.0273, over 1193501.83 frames. ], batch size: 64, lr: 5.04e-03, grad_scale: 8.0 +2023-03-21 06:43:42,928 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.788e+02 2.029e+02 2.568e+02 9.373e+02, threshold=4.059e+02, percent-clipped=4.0 +2023-03-21 06:43:44,539 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 06:43:58,676 INFO [train.py:901] (1/2) Epoch 33, batch 400, loss[loss=0.1173, simple_loss=0.1971, pruned_loss=0.0187, over 7367.00 frames. ], tot_loss[loss=0.135, simple_loss=0.2156, pruned_loss=0.02717, over 1250396.98 frames. ], batch size: 44, lr: 5.03e-03, grad_scale: 8.0 +2023-03-21 06:43:59,462 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-03-21 06:44:07,213 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90786.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:44:08,297 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2385, 2.8512, 2.0318, 3.4105, 3.1626, 3.3761, 2.9169, 2.9329], + device='cuda:1'), covar=tensor([0.2019, 0.0928, 0.3849, 0.0588, 0.0246, 0.0286, 0.0386, 0.0401], + device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0233, 0.0252, 0.0255, 0.0189, 0.0187, 0.0208, 0.0219], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 06:44:22,885 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90815.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:44:24,763 INFO [train.py:901] (1/2) Epoch 33, batch 450, loss[loss=0.1375, simple_loss=0.2253, pruned_loss=0.02487, over 7358.00 frames. ], tot_loss[loss=0.1349, simple_loss=0.2152, pruned_loss=0.02728, over 1290818.97 frames. ], batch size: 63, lr: 5.03e-03, grad_scale: 8.0 +2023-03-21 06:44:27,032 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.21 vs. limit=5.0 +2023-03-21 06:44:28,704 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 06:44:29,221 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 06:44:32,287 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=90834.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:44:34,677 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.801e+02 2.162e+02 2.490e+02 7.206e+02, threshold=4.324e+02, percent-clipped=3.0 +2023-03-21 06:44:39,895 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90849.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:44:41,953 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90852.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:44:50,278 INFO [train.py:901] (1/2) Epoch 33, batch 500, loss[loss=0.1606, simple_loss=0.2413, pruned_loss=0.03994, over 7236.00 frames. ], tot_loss[loss=0.1346, simple_loss=0.2147, pruned_loss=0.02723, over 1323090.79 frames. ], batch size: 55, lr: 5.03e-03, grad_scale: 8.0 +2023-03-21 06:45:00,479 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90888.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:45:01,432 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 06:45:02,949 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 06:45:03,471 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 06:45:05,440 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 06:45:10,434 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 06:45:11,589 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90910.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:45:15,844 INFO [train.py:901] (1/2) Epoch 33, batch 550, loss[loss=0.1385, simple_loss=0.22, pruned_loss=0.02857, over 7312.00 frames. ], tot_loss[loss=0.1345, simple_loss=0.215, pruned_loss=0.02704, over 1350480.85 frames. ], batch size: 49, lr: 5.03e-03, grad_scale: 8.0 +2023-03-21 06:45:17,485 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7899, 2.6859, 2.4301, 3.7236, 1.8588, 3.7686, 1.7216, 3.2847], + device='cuda:1'), covar=tensor([0.0177, 0.1197, 0.1679, 0.0157, 0.3777, 0.0231, 0.1112, 0.0391], + device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0250, 0.0263, 0.0202, 0.0251, 0.0210, 0.0232, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 06:45:20,873 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 06:45:25,069 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=90936.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:45:26,467 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.416e+02 1.907e+02 2.191e+02 2.571e+02 3.515e+02, threshold=4.383e+02, percent-clipped=0.0 +2023-03-21 06:45:30,016 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 06:45:32,974 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 06:45:40,040 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 06:45:42,044 INFO [train.py:901] (1/2) Epoch 33, batch 600, loss[loss=0.1404, simple_loss=0.2264, pruned_loss=0.02722, over 7301.00 frames. ], tot_loss[loss=0.135, simple_loss=0.2153, pruned_loss=0.02731, over 1371375.97 frames. ], batch size: 80, lr: 5.03e-03, grad_scale: 8.0 +2023-03-21 06:45:56,184 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 06:46:01,718 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91007.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:46:05,686 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 06:46:07,659 INFO [train.py:901] (1/2) Epoch 33, batch 650, loss[loss=0.1227, simple_loss=0.2021, pruned_loss=0.02165, over 7341.00 frames. ], tot_loss[loss=0.1346, simple_loss=0.2149, pruned_loss=0.0272, over 1387376.11 frames. ], batch size: 44, lr: 5.03e-03, grad_scale: 8.0 +2023-03-21 06:46:18,395 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.079e+02 1.852e+02 2.052e+02 2.347e+02 4.499e+02, threshold=4.105e+02, percent-clipped=1.0 +2023-03-21 06:46:24,353 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 06:46:27,031 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=91055.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:46:32,955 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 06:46:33,927 INFO [train.py:901] (1/2) Epoch 33, batch 700, loss[loss=0.1529, simple_loss=0.2351, pruned_loss=0.0353, over 7367.00 frames. ], tot_loss[loss=0.1347, simple_loss=0.2149, pruned_loss=0.02728, over 1399133.92 frames. ], batch size: 63, lr: 5.03e-03, grad_scale: 8.0 +2023-03-21 06:46:43,010 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4360, 1.3784, 1.6467, 1.9685, 1.7246, 1.9664, 1.3412, 1.8886], + device='cuda:1'), covar=tensor([0.2866, 0.4469, 0.1599, 0.1000, 0.1488, 0.1825, 0.1716, 0.1799], + device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0072, 0.0061, 0.0058, 0.0057, 0.0057, 0.0096, 0.0059], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:46:43,485 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8291, 3.8873, 3.0717, 3.4134, 2.7108, 2.1455, 1.7902, 3.8342], + device='cuda:1'), covar=tensor([0.0050, 0.0038, 0.0153, 0.0062, 0.0190, 0.0538, 0.0654, 0.0051], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0085, 0.0107, 0.0091, 0.0121, 0.0128, 0.0126, 0.0099], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 06:46:45,669 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 +2023-03-21 06:46:57,018 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 06:46:57,492 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 06:46:57,583 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91115.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:46:58,174 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1768, 2.5461, 1.8946, 2.8883, 2.7739, 2.9833, 2.6439, 2.6285], + device='cuda:1'), covar=tensor([0.1941, 0.0937, 0.3489, 0.0741, 0.0255, 0.0276, 0.0357, 0.0407], + device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0233, 0.0251, 0.0256, 0.0190, 0.0189, 0.0209, 0.0219], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 06:46:59,483 INFO [train.py:901] (1/2) Epoch 33, batch 750, loss[loss=0.104, simple_loss=0.1852, pruned_loss=0.01135, over 7147.00 frames. ], tot_loss[loss=0.1347, simple_loss=0.2152, pruned_loss=0.02704, over 1407799.95 frames. ], batch size: 41, lr: 5.02e-03, grad_scale: 8.0 +2023-03-21 06:47:10,048 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.773e+02 2.102e+02 2.526e+02 4.184e+02, threshold=4.205e+02, percent-clipped=1.0 +2023-03-21 06:47:11,636 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 06:47:16,072 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 06:47:16,635 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91152.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:47:21,581 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 06:47:22,085 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=91163.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:47:23,020 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 06:47:25,013 INFO [train.py:901] (1/2) Epoch 33, batch 800, loss[loss=0.1213, simple_loss=0.2047, pruned_loss=0.01898, over 7343.00 frames. ], tot_loss[loss=0.1348, simple_loss=0.2154, pruned_loss=0.02706, over 1417134.83 frames. ], batch size: 44, lr: 5.02e-03, grad_scale: 8.0 +2023-03-21 06:47:32,546 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 06:47:41,706 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=91200.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:47:44,247 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91205.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:47:51,133 INFO [train.py:901] (1/2) Epoch 33, batch 850, loss[loss=0.1373, simple_loss=0.223, pruned_loss=0.02585, over 7358.00 frames. ], tot_loss[loss=0.1344, simple_loss=0.2149, pruned_loss=0.02698, over 1418873.11 frames. ], batch size: 63, lr: 5.02e-03, grad_scale: 8.0 +2023-03-21 06:47:54,332 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 06:47:54,341 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 06:47:59,411 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 06:48:01,501 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7644, 3.8414, 3.0812, 3.3865, 2.7829, 2.0994, 1.9086, 3.8153], + device='cuda:1'), covar=tensor([0.0085, 0.0066, 0.0208, 0.0091, 0.0247, 0.0682, 0.0713, 0.0084], + device='cuda:1'), in_proj_covar=tensor([0.0108, 0.0086, 0.0107, 0.0092, 0.0122, 0.0129, 0.0127, 0.0100], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 06:48:01,844 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.326e+02 1.733e+02 2.042e+02 2.278e+02 4.031e+02, threshold=4.085e+02, percent-clipped=0.0 +2023-03-21 06:48:02,894 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 06:48:16,899 INFO [train.py:901] (1/2) Epoch 33, batch 900, loss[loss=0.1161, simple_loss=0.198, pruned_loss=0.01714, over 7162.00 frames. ], tot_loss[loss=0.1342, simple_loss=0.215, pruned_loss=0.02674, over 1425580.12 frames. ], batch size: 41, lr: 5.02e-03, grad_scale: 8.0 +2023-03-21 06:48:21,949 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8483, 4.0036, 3.7601, 4.0122, 3.6858, 3.8730, 4.1909, 4.2267], + device='cuda:1'), covar=tensor([0.0202, 0.0141, 0.0198, 0.0160, 0.0262, 0.0287, 0.0280, 0.0182], + device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0120, 0.0113, 0.0116, 0.0107, 0.0096, 0.0095, 0.0092], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:48:25,085 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91285.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:48:40,888 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 06:48:41,766 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 06:48:43,383 INFO [train.py:901] (1/2) Epoch 33, batch 950, loss[loss=0.1257, simple_loss=0.2109, pruned_loss=0.02023, over 7264.00 frames. ], tot_loss[loss=0.1347, simple_loss=0.2151, pruned_loss=0.02721, over 1428169.40 frames. ], batch size: 89, lr: 5.02e-03, grad_scale: 8.0 +2023-03-21 06:48:43,555 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2474, 2.6091, 1.9368, 2.9152, 2.8120, 2.9869, 2.6939, 2.7497], + device='cuda:1'), covar=tensor([0.2254, 0.0956, 0.3742, 0.0714, 0.0311, 0.0252, 0.0384, 0.0399], + device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0234, 0.0252, 0.0256, 0.0190, 0.0190, 0.0210, 0.0221], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 06:48:45,575 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5496, 2.0073, 1.4878, 1.7009, 1.9127, 1.4792, 1.6186, 1.1938], + device='cuda:1'), covar=tensor([0.0175, 0.0114, 0.0367, 0.0167, 0.0096, 0.0182, 0.0172, 0.0222], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0032, 0.0031, 0.0032, 0.0031, 0.0030, 0.0034, 0.0041], + device='cuda:1'), out_proj_covar=tensor([3.9201e-05, 3.5685e-05, 3.5847e-05, 3.6109e-05, 3.4867e-05, 3.3496e-05, + 3.8374e-05, 4.5539e-05], device='cuda:1') +2023-03-21 06:48:53,316 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.253e+02 1.750e+02 2.027e+02 2.415e+02 4.015e+02, threshold=4.054e+02, percent-clipped=0.0 +2023-03-21 06:48:56,996 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91346.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:49:03,959 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 06:49:08,354 INFO [train.py:901] (1/2) Epoch 33, batch 1000, loss[loss=0.1426, simple_loss=0.2232, pruned_loss=0.03096, over 7259.00 frames. ], tot_loss[loss=0.1344, simple_loss=0.2148, pruned_loss=0.02705, over 1432378.15 frames. ], batch size: 55, lr: 5.02e-03, grad_scale: 8.0 +2023-03-21 06:49:19,143 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 +2023-03-21 06:49:24,291 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.2758, 1.5001, 1.3008, 1.3079, 1.4459, 1.4186, 1.3450, 1.2075], + device='cuda:1'), covar=tensor([0.0147, 0.0103, 0.0195, 0.0118, 0.0104, 0.0116, 0.0107, 0.0128], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0032, 0.0032, 0.0033, 0.0031, 0.0030, 0.0034, 0.0041], + device='cuda:1'), out_proj_covar=tensor([3.9364e-05, 3.5725e-05, 3.5956e-05, 3.6209e-05, 3.4970e-05, 3.3745e-05, + 3.8192e-05, 4.5569e-05], device='cuda:1') +2023-03-21 06:49:24,625 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 06:49:34,750 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-03-21 06:49:34,938 INFO [train.py:901] (1/2) Epoch 33, batch 1050, loss[loss=0.132, simple_loss=0.2101, pruned_loss=0.02699, over 7244.00 frames. ], tot_loss[loss=0.1348, simple_loss=0.2153, pruned_loss=0.02721, over 1434201.76 frames. ], batch size: 45, lr: 5.02e-03, grad_scale: 8.0 +2023-03-21 06:49:36,105 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6043, 1.5360, 1.9046, 2.2476, 2.0766, 2.2573, 1.8810, 2.1348], + device='cuda:1'), covar=tensor([0.1818, 0.5991, 0.2174, 0.0801, 0.3000, 0.1550, 0.1656, 0.3820], + device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0072, 0.0061, 0.0057, 0.0057, 0.0057, 0.0095, 0.0059], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:49:44,965 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.697e+02 1.949e+02 2.264e+02 4.726e+02, threshold=3.897e+02, percent-clipped=1.0 +2023-03-21 06:49:45,575 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6869, 4.1968, 4.3612, 4.5169, 4.2794, 4.2954, 4.6076, 4.2009], + device='cuda:1'), covar=tensor([0.0128, 0.0131, 0.0130, 0.0107, 0.0419, 0.0094, 0.0117, 0.0125], + device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0099, 0.0099, 0.0088, 0.0173, 0.0105, 0.0104, 0.0109], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:49:45,995 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 06:49:46,254 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-03-21 06:49:49,995 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 06:49:59,021 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-03-21 06:50:00,704 INFO [train.py:901] (1/2) Epoch 33, batch 1100, loss[loss=0.1443, simple_loss=0.2179, pruned_loss=0.03535, over 7241.00 frames. ], tot_loss[loss=0.1342, simple_loss=0.2143, pruned_loss=0.02708, over 1435401.67 frames. ], batch size: 45, lr: 5.01e-03, grad_scale: 8.0 +2023-03-21 06:50:19,557 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91505.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:50:19,983 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 06:50:20,767 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 +2023-03-21 06:50:20,916 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 06:50:26,425 INFO [train.py:901] (1/2) Epoch 33, batch 1150, loss[loss=0.1398, simple_loss=0.2234, pruned_loss=0.02812, over 7309.00 frames. ], tot_loss[loss=0.1344, simple_loss=0.2148, pruned_loss=0.02697, over 1439167.12 frames. ], batch size: 59, lr: 5.01e-03, grad_scale: 8.0 +2023-03-21 06:50:33,000 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 06:50:33,504 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 06:50:36,492 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.779e+02 2.098e+02 2.565e+02 4.143e+02, threshold=4.196e+02, percent-clipped=2.0 +2023-03-21 06:50:38,181 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91542.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 06:50:43,761 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1017, 4.1030, 3.5824, 3.7170, 2.9669, 2.3370, 2.0159, 4.1507], + device='cuda:1'), covar=tensor([0.0051, 0.0062, 0.0121, 0.0065, 0.0173, 0.0519, 0.0599, 0.0047], + device='cuda:1'), in_proj_covar=tensor([0.0107, 0.0086, 0.0108, 0.0091, 0.0122, 0.0129, 0.0126, 0.0100], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 06:50:44,203 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=91553.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:50:50,980 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-21 06:50:52,782 INFO [train.py:901] (1/2) Epoch 33, batch 1200, loss[loss=0.1415, simple_loss=0.224, pruned_loss=0.02956, over 7351.00 frames. ], tot_loss[loss=0.1347, simple_loss=0.2153, pruned_loss=0.02702, over 1439951.30 frames. ], batch size: 54, lr: 5.01e-03, grad_scale: 8.0 +2023-03-21 06:51:05,855 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 06:51:07,196 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-03-21 06:51:10,350 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91603.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:51:18,290 INFO [train.py:901] (1/2) Epoch 33, batch 1250, loss[loss=0.136, simple_loss=0.2175, pruned_loss=0.02727, over 7231.00 frames. ], tot_loss[loss=0.1345, simple_loss=0.2152, pruned_loss=0.02694, over 1440078.85 frames. ], batch size: 93, lr: 5.01e-03, grad_scale: 8.0 +2023-03-21 06:51:28,865 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.321e+02 1.782e+02 2.023e+02 2.411e+02 3.867e+02, threshold=4.047e+02, percent-clipped=0.0 +2023-03-21 06:51:29,943 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91641.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:51:30,893 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 06:51:35,000 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 06:51:36,471 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 06:51:39,196 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8250, 3.3361, 3.0347, 3.3290, 3.2562, 3.0689, 3.0992, 3.1340], + device='cuda:1'), covar=tensor([0.0822, 0.1318, 0.0536, 0.1125, 0.1751, 0.0630, 0.1454, 0.0751], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0056, 0.0064, 0.0055, 0.0053, 0.0057, 0.0055, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:51:44,524 INFO [train.py:901] (1/2) Epoch 33, batch 1300, loss[loss=0.1364, simple_loss=0.2156, pruned_loss=0.02862, over 7123.00 frames. ], tot_loss[loss=0.1337, simple_loss=0.2144, pruned_loss=0.02654, over 1438730.48 frames. ], batch size: 41, lr: 5.01e-03, grad_scale: 8.0 +2023-03-21 06:51:46,123 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7289, 5.1488, 5.2085, 5.1825, 4.8967, 4.6282, 5.2115, 4.9957], + device='cuda:1'), covar=tensor([0.0397, 0.0408, 0.0349, 0.0443, 0.0395, 0.0383, 0.0353, 0.0521], + device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0248, 0.0188, 0.0192, 0.0150, 0.0221, 0.0196, 0.0145], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:51:50,193 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4048, 3.2491, 3.1705, 3.2165, 2.8095, 2.7828, 3.2721, 2.5094], + device='cuda:1'), covar=tensor([0.0410, 0.0541, 0.0695, 0.0548, 0.0811, 0.1111, 0.0766, 0.2045], + device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0339, 0.0273, 0.0357, 0.0293, 0.0290, 0.0346, 0.0258], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:51:59,653 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 06:52:02,136 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 06:52:05,648 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 06:52:09,566 INFO [train.py:901] (1/2) Epoch 33, batch 1350, loss[loss=0.1446, simple_loss=0.2251, pruned_loss=0.0321, over 7221.00 frames. ], tot_loss[loss=0.1337, simple_loss=0.2147, pruned_loss=0.02636, over 1441051.59 frames. ], batch size: 93, lr: 5.01e-03, grad_scale: 8.0 +2023-03-21 06:52:15,701 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 06:52:15,857 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1711, 2.8589, 1.9731, 2.9311, 3.1547, 3.0906, 2.7821, 2.6705], + device='cuda:1'), covar=tensor([0.2224, 0.0967, 0.3661, 0.0771, 0.0252, 0.0205, 0.0349, 0.0473], + device='cuda:1'), in_proj_covar=tensor([0.0258, 0.0235, 0.0252, 0.0258, 0.0190, 0.0190, 0.0211, 0.0222], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 06:52:17,284 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9758, 4.0615, 3.7969, 4.0580, 3.7701, 3.9642, 4.2699, 4.3073], + device='cuda:1'), covar=tensor([0.0191, 0.0150, 0.0235, 0.0158, 0.0344, 0.0256, 0.0281, 0.0190], + device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0124, 0.0116, 0.0119, 0.0110, 0.0100, 0.0097, 0.0095], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:52:20,215 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.841e+02 2.177e+02 2.731e+02 4.571e+02, threshold=4.353e+02, percent-clipped=2.0 +2023-03-21 06:52:36,005 INFO [train.py:901] (1/2) Epoch 33, batch 1400, loss[loss=0.1458, simple_loss=0.2266, pruned_loss=0.03251, over 7286.00 frames. ], tot_loss[loss=0.134, simple_loss=0.2148, pruned_loss=0.02665, over 1439533.87 frames. ], batch size: 66, lr: 5.01e-03, grad_scale: 8.0 +2023-03-21 06:52:48,610 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 06:53:02,051 INFO [train.py:901] (1/2) Epoch 33, batch 1450, loss[loss=0.1377, simple_loss=0.222, pruned_loss=0.02672, over 7310.00 frames. ], tot_loss[loss=0.1338, simple_loss=0.2147, pruned_loss=0.02645, over 1441124.45 frames. ], batch size: 83, lr: 5.01e-03, grad_scale: 8.0 +2023-03-21 06:53:12,889 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.866e+02 2.159e+02 2.546e+02 4.457e+02, threshold=4.319e+02, percent-clipped=1.0 +2023-03-21 06:53:13,417 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 06:53:27,928 INFO [train.py:901] (1/2) Epoch 33, batch 1500, loss[loss=0.1275, simple_loss=0.2069, pruned_loss=0.024, over 7345.00 frames. ], tot_loss[loss=0.1342, simple_loss=0.2148, pruned_loss=0.02677, over 1441051.61 frames. ], batch size: 54, lr: 5.00e-03, grad_scale: 8.0 +2023-03-21 06:53:30,512 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 06:53:37,595 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1334, 2.8811, 2.0994, 3.4411, 2.4057, 2.8680, 1.4395, 2.2610], + device='cuda:1'), covar=tensor([0.0516, 0.0943, 0.2598, 0.0723, 0.0530, 0.0645, 0.3752, 0.1875], + device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0255, 0.0281, 0.0268, 0.0267, 0.0265, 0.0236, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 06:53:39,032 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91891.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:53:43,111 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91898.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 06:53:54,182 INFO [train.py:901] (1/2) Epoch 33, batch 1550, loss[loss=0.1204, simple_loss=0.2064, pruned_loss=0.01722, over 7335.00 frames. ], tot_loss[loss=0.134, simple_loss=0.2148, pruned_loss=0.02659, over 1441948.50 frames. ], batch size: 75, lr: 5.00e-03, grad_scale: 8.0 +2023-03-21 06:53:55,738 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 06:54:04,240 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.339e+02 1.806e+02 2.097e+02 2.393e+02 4.199e+02, threshold=4.194e+02, percent-clipped=0.0 +2023-03-21 06:54:05,390 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91941.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:54:10,980 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91952.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:54:12,458 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91955.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:54:19,275 INFO [train.py:901] (1/2) Epoch 33, batch 1600, loss[loss=0.1353, simple_loss=0.2133, pruned_loss=0.02866, over 7232.00 frames. ], tot_loss[loss=0.134, simple_loss=0.215, pruned_loss=0.02647, over 1442143.55 frames. ], batch size: 45, lr: 5.00e-03, grad_scale: 8.0 +2023-03-21 06:54:24,943 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 06:54:25,431 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 06:54:25,558 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91981.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 06:54:29,015 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 06:54:30,084 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=91989.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:54:32,610 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0015, 3.2948, 3.8741, 4.0212, 3.9622, 4.0140, 4.0647, 3.9313], + device='cuda:1'), covar=tensor([0.0029, 0.0107, 0.0032, 0.0034, 0.0031, 0.0032, 0.0034, 0.0045], + device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0066, 0.0054, 0.0053, 0.0052, 0.0058, 0.0047, 0.0073], + device='cuda:1'), out_proj_covar=tensor([7.9597e-05, 1.3901e-04, 1.0218e-04, 9.4775e-05, 9.2812e-05, 1.0484e-04, + 9.2809e-05, 1.3992e-04], device='cuda:1') +2023-03-21 06:54:43,097 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 06:54:47,073 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 06:54:48,278 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92016.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:54:49,718 INFO [train.py:901] (1/2) Epoch 33, batch 1650, loss[loss=0.1505, simple_loss=0.2268, pruned_loss=0.03707, over 7328.00 frames. ], tot_loss[loss=0.1341, simple_loss=0.2152, pruned_loss=0.02657, over 1440122.82 frames. ], batch size: 75, lr: 5.00e-03, grad_scale: 8.0 +2023-03-21 06:54:55,362 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 06:54:59,841 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.781e+02 2.039e+02 2.347e+02 4.306e+02, threshold=4.078e+02, percent-clipped=2.0 +2023-03-21 06:55:01,519 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92042.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 06:55:11,419 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 06:55:14,852 INFO [train.py:901] (1/2) Epoch 33, batch 1700, loss[loss=0.1322, simple_loss=0.2164, pruned_loss=0.02402, over 7286.00 frames. ], tot_loss[loss=0.1349, simple_loss=0.2161, pruned_loss=0.02686, over 1441949.45 frames. ], batch size: 57, lr: 5.00e-03, grad_scale: 8.0 +2023-03-21 06:55:15,899 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 06:55:27,655 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 06:55:41,160 INFO [train.py:901] (1/2) Epoch 33, batch 1750, loss[loss=0.1328, simple_loss=0.2147, pruned_loss=0.02543, over 7272.00 frames. ], tot_loss[loss=0.1349, simple_loss=0.2163, pruned_loss=0.02676, over 1443746.83 frames. ], batch size: 57, lr: 5.00e-03, grad_scale: 8.0 +2023-03-21 06:55:51,185 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.689e+02 2.025e+02 2.316e+02 3.712e+02, threshold=4.050e+02, percent-clipped=0.0 +2023-03-21 06:55:51,213 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 06:55:52,248 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 06:56:07,477 INFO [train.py:901] (1/2) Epoch 33, batch 1800, loss[loss=0.144, simple_loss=0.2215, pruned_loss=0.03327, over 7338.00 frames. ], tot_loss[loss=0.1347, simple_loss=0.2159, pruned_loss=0.02675, over 1443929.37 frames. ], batch size: 54, lr: 5.00e-03, grad_scale: 8.0 +2023-03-21 06:56:08,846 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 +2023-03-21 06:56:15,449 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 06:56:22,105 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92198.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 06:56:27,395 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7038, 3.4338, 3.4767, 3.4130, 3.3402, 3.2506, 3.6009, 3.2411], + device='cuda:1'), covar=tensor([0.0135, 0.0201, 0.0133, 0.0212, 0.0486, 0.0130, 0.0162, 0.0203], + device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0100, 0.0099, 0.0088, 0.0174, 0.0104, 0.0103, 0.0108], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:56:28,825 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 06:56:32,726 INFO [train.py:901] (1/2) Epoch 33, batch 1850, loss[loss=0.1396, simple_loss=0.2145, pruned_loss=0.03234, over 7247.00 frames. ], tot_loss[loss=0.1348, simple_loss=0.2158, pruned_loss=0.02689, over 1442923.00 frames. ], batch size: 47, lr: 4.99e-03, grad_scale: 8.0 +2023-03-21 06:56:38,718 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 06:56:42,787 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.249e+02 1.760e+02 1.984e+02 2.426e+02 4.965e+02, threshold=3.969e+02, percent-clipped=2.0 +2023-03-21 06:56:45,542 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 +2023-03-21 06:56:46,345 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=92246.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 06:56:46,806 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92247.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:56:48,371 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9979, 4.3874, 4.4157, 4.3707, 4.4059, 4.0090, 4.4558, 4.3642], + device='cuda:1'), covar=tensor([0.0471, 0.0402, 0.0382, 0.0498, 0.0265, 0.0398, 0.0324, 0.0413], + device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0248, 0.0189, 0.0190, 0.0150, 0.0221, 0.0195, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:56:56,506 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 06:56:59,088 INFO [train.py:901] (1/2) Epoch 33, batch 1900, loss[loss=0.1349, simple_loss=0.2073, pruned_loss=0.0312, over 7272.00 frames. ], tot_loss[loss=0.135, simple_loss=0.2161, pruned_loss=0.02697, over 1445011.37 frames. ], batch size: 47, lr: 4.99e-03, grad_scale: 8.0 +2023-03-21 06:57:20,212 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92311.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:57:21,153 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 06:57:23,307 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3011, 2.6490, 2.0921, 3.0243, 2.9212, 3.0481, 2.8685, 2.7193], + device='cuda:1'), covar=tensor([0.1911, 0.0941, 0.3460, 0.0677, 0.0246, 0.0198, 0.0282, 0.0363], + device='cuda:1'), in_proj_covar=tensor([0.0254, 0.0233, 0.0250, 0.0256, 0.0191, 0.0189, 0.0208, 0.0220], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 06:57:24,136 INFO [train.py:901] (1/2) Epoch 33, batch 1950, loss[loss=0.1289, simple_loss=0.2096, pruned_loss=0.02408, over 7276.00 frames. ], tot_loss[loss=0.1342, simple_loss=0.2152, pruned_loss=0.02658, over 1444084.18 frames. ], batch size: 47, lr: 4.99e-03, grad_scale: 8.0 +2023-03-21 06:57:31,741 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 06:57:33,919 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92337.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 06:57:34,756 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.334e+02 1.739e+02 2.064e+02 2.518e+02 6.766e+02, threshold=4.128e+02, percent-clipped=1.0 +2023-03-21 06:57:36,867 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 06:57:37,346 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 06:57:50,556 INFO [train.py:901] (1/2) Epoch 33, batch 2000, loss[loss=0.1413, simple_loss=0.2237, pruned_loss=0.02948, over 7342.00 frames. ], tot_loss[loss=0.1346, simple_loss=0.2157, pruned_loss=0.02677, over 1444645.59 frames. ], batch size: 59, lr: 4.99e-03, grad_scale: 8.0 +2023-03-21 06:57:54,120 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 06:58:04,676 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 06:58:10,645 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-21 06:58:13,440 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 06:58:14,526 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7516, 5.2146, 5.2941, 5.1867, 4.9727, 4.7196, 5.3328, 5.1243], + device='cuda:1'), covar=tensor([0.0378, 0.0351, 0.0329, 0.0447, 0.0308, 0.0330, 0.0251, 0.0397], + device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0248, 0.0189, 0.0191, 0.0150, 0.0221, 0.0197, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 06:58:15,977 INFO [train.py:901] (1/2) Epoch 33, batch 2050, loss[loss=0.1398, simple_loss=0.2192, pruned_loss=0.03023, over 7319.00 frames. ], tot_loss[loss=0.1352, simple_loss=0.2163, pruned_loss=0.02701, over 1445696.83 frames. ], batch size: 49, lr: 4.99e-03, grad_scale: 8.0 +2023-03-21 06:58:27,242 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.309e+02 1.918e+02 2.279e+02 2.660e+02 4.216e+02, threshold=4.557e+02, percent-clipped=1.0 +2023-03-21 06:58:42,295 INFO [train.py:901] (1/2) Epoch 33, batch 2100, loss[loss=0.1401, simple_loss=0.2132, pruned_loss=0.03348, over 7341.00 frames. ], tot_loss[loss=0.1346, simple_loss=0.2155, pruned_loss=0.02681, over 1444927.24 frames. ], batch size: 51, lr: 4.99e-03, grad_scale: 8.0 +2023-03-21 06:58:46,276 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 06:58:49,273 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 06:58:53,897 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.8037, 4.4108, 4.1679, 4.7566, 4.6110, 4.7043, 4.2008, 4.3845], + device='cuda:1'), covar=tensor([0.0895, 0.2444, 0.2228, 0.1069, 0.0951, 0.1177, 0.0746, 0.1243], + device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0379, 0.0285, 0.0296, 0.0219, 0.0358, 0.0216, 0.0263], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:59:08,982 INFO [train.py:901] (1/2) Epoch 33, batch 2150, loss[loss=0.138, simple_loss=0.2173, pruned_loss=0.02929, over 7262.00 frames. ], tot_loss[loss=0.1343, simple_loss=0.2151, pruned_loss=0.02677, over 1445335.90 frames. ], batch size: 47, lr: 4.99e-03, grad_scale: 8.0 +2023-03-21 06:59:19,166 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.208e+02 1.797e+02 1.982e+02 2.424e+02 4.305e+02, threshold=3.964e+02, percent-clipped=0.0 +2023-03-21 06:59:19,815 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2937, 3.8560, 4.0533, 3.9926, 3.8961, 3.8508, 4.0886, 3.7465], + device='cuda:1'), covar=tensor([0.0116, 0.0193, 0.0100, 0.0150, 0.0384, 0.0114, 0.0161, 0.0167], + device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0098, 0.0096, 0.0086, 0.0171, 0.0102, 0.0102, 0.0106], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 06:59:23,377 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92547.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:59:34,084 INFO [train.py:901] (1/2) Epoch 33, batch 2200, loss[loss=0.1404, simple_loss=0.2243, pruned_loss=0.02824, over 7294.00 frames. ], tot_loss[loss=0.1345, simple_loss=0.2152, pruned_loss=0.02691, over 1441555.14 frames. ], batch size: 68, lr: 4.98e-03, grad_scale: 8.0 +2023-03-21 06:59:35,609 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 06:59:48,009 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=92595.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 06:59:56,949 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92611.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:00:00,849 INFO [train.py:901] (1/2) Epoch 33, batch 2250, loss[loss=0.143, simple_loss=0.2302, pruned_loss=0.02789, over 6656.00 frames. ], tot_loss[loss=0.1349, simple_loss=0.2157, pruned_loss=0.02703, over 1439513.06 frames. ], batch size: 106, lr: 4.98e-03, grad_scale: 8.0 +2023-03-21 07:00:03,069 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8377, 3.0876, 2.5539, 2.8183, 3.0810, 2.5598, 3.0217, 2.8979], + device='cuda:1'), covar=tensor([0.0684, 0.0823, 0.1362, 0.1322, 0.1238, 0.0832, 0.0762, 0.1266], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0056, 0.0065, 0.0056, 0.0053, 0.0058, 0.0055, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:00:04,078 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92625.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:00:09,050 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 07:00:09,516 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 07:00:10,096 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92637.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 07:00:10,963 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.180e+02 1.747e+02 2.128e+02 2.442e+02 8.863e+02, threshold=4.255e+02, percent-clipped=5.0 +2023-03-21 07:00:11,576 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.8784, 4.4225, 4.3149, 4.8060, 4.7267, 4.7351, 4.2334, 4.4050], + device='cuda:1'), covar=tensor([0.0900, 0.2672, 0.2407, 0.1132, 0.0927, 0.1376, 0.0837, 0.1240], + device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0378, 0.0284, 0.0297, 0.0219, 0.0357, 0.0215, 0.0265], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:00:13,996 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 +2023-03-21 07:00:21,213 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=92659.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:00:21,677 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 07:00:23,775 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9103, 3.8254, 3.2037, 3.6001, 3.1249, 2.0815, 1.7859, 4.0170], + device='cuda:1'), covar=tensor([0.0085, 0.0108, 0.0224, 0.0091, 0.0210, 0.0764, 0.0825, 0.0069], + device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0089, 0.0110, 0.0095, 0.0125, 0.0132, 0.0129, 0.0102], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 07:00:26,182 INFO [train.py:901] (1/2) Epoch 33, batch 2300, loss[loss=0.1479, simple_loss=0.2291, pruned_loss=0.03333, over 7214.00 frames. ], tot_loss[loss=0.1355, simple_loss=0.2167, pruned_loss=0.02713, over 1441508.04 frames. ], batch size: 93, lr: 4.98e-03, grad_scale: 16.0 +2023-03-21 07:00:34,719 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-21 07:00:35,522 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=92685.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 07:00:36,047 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92686.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:00:52,479 INFO [train.py:901] (1/2) Epoch 33, batch 2350, loss[loss=0.1286, simple_loss=0.214, pruned_loss=0.02162, over 7319.00 frames. ], tot_loss[loss=0.1341, simple_loss=0.2153, pruned_loss=0.02641, over 1442179.86 frames. ], batch size: 59, lr: 4.98e-03, grad_scale: 16.0 +2023-03-21 07:00:53,119 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92720.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:01:02,574 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.641e+02 1.954e+02 2.231e+02 4.206e+02, threshold=3.909e+02, percent-clipped=0.0 +2023-03-21 07:01:02,745 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7481, 4.4893, 4.5285, 4.2315, 4.1528, 3.0675, 2.9063, 4.8443], + device='cuda:1'), covar=tensor([0.0044, 0.0097, 0.0064, 0.0061, 0.0086, 0.0438, 0.0447, 0.0047], + device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0089, 0.0109, 0.0095, 0.0124, 0.0132, 0.0129, 0.0102], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 07:01:10,239 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 07:01:16,955 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 07:01:18,387 INFO [train.py:901] (1/2) Epoch 33, batch 2400, loss[loss=0.1267, simple_loss=0.2109, pruned_loss=0.02123, over 7273.00 frames. ], tot_loss[loss=0.1336, simple_loss=0.2146, pruned_loss=0.0263, over 1440651.56 frames. ], batch size: 66, lr: 4.98e-03, grad_scale: 16.0 +2023-03-21 07:01:25,419 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92781.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:01:28,378 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 07:01:31,893 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 07:01:44,635 INFO [train.py:901] (1/2) Epoch 33, batch 2450, loss[loss=0.1459, simple_loss=0.233, pruned_loss=0.02935, over 7307.00 frames. ], tot_loss[loss=0.1338, simple_loss=0.2148, pruned_loss=0.02636, over 1440514.60 frames. ], batch size: 86, lr: 4.98e-03, grad_scale: 16.0 +2023-03-21 07:01:48,837 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1293, 4.5917, 4.6139, 4.5576, 4.5440, 4.1469, 4.6727, 4.5206], + device='cuda:1'), covar=tensor([0.0482, 0.0398, 0.0382, 0.0539, 0.0304, 0.0363, 0.0309, 0.0363], + device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0246, 0.0189, 0.0189, 0.0149, 0.0218, 0.0196, 0.0143], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:01:50,649 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-03-21 07:01:50,941 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92831.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:01:54,888 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.351e+02 1.774e+02 2.037e+02 2.369e+02 4.194e+02, threshold=4.073e+02, percent-clipped=1.0 +2023-03-21 07:01:58,978 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 07:02:11,265 INFO [train.py:901] (1/2) Epoch 33, batch 2500, loss[loss=0.1334, simple_loss=0.2172, pruned_loss=0.02479, over 7313.00 frames. ], tot_loss[loss=0.1345, simple_loss=0.2155, pruned_loss=0.02671, over 1439415.64 frames. ], batch size: 83, lr: 4.98e-03, grad_scale: 16.0 +2023-03-21 07:02:23,012 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92892.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:02:24,860 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 07:02:26,487 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92899.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:02:31,230 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 +2023-03-21 07:02:36,323 INFO [train.py:901] (1/2) Epoch 33, batch 2550, loss[loss=0.1355, simple_loss=0.2165, pruned_loss=0.0272, over 7354.00 frames. ], tot_loss[loss=0.1342, simple_loss=0.2151, pruned_loss=0.02667, over 1438748.28 frames. ], batch size: 73, lr: 4.98e-03, grad_scale: 8.0 +2023-03-21 07:02:42,444 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92931.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:02:46,260 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 +2023-03-21 07:02:46,899 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.311e+02 1.757e+02 2.040e+02 2.452e+02 4.787e+02, threshold=4.080e+02, percent-clipped=2.0 +2023-03-21 07:02:52,854 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5777, 2.2477, 2.7700, 2.6308, 2.6858, 2.4962, 2.1233, 2.6557], + device='cuda:1'), covar=tensor([0.1213, 0.0963, 0.0971, 0.1231, 0.1017, 0.1099, 0.2364, 0.1261], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0065, 0.0049, 0.0049, 0.0048, 0.0048, 0.0066, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 07:02:56,384 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1403, 3.3250, 2.3285, 3.7197, 2.6184, 3.2607, 1.6449, 2.2803], + device='cuda:1'), covar=tensor([0.0405, 0.0641, 0.2322, 0.0590, 0.0462, 0.0625, 0.3264, 0.1773], + device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0255, 0.0282, 0.0270, 0.0270, 0.0267, 0.0238, 0.0263], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 07:02:58,344 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92960.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:03:02,611 INFO [train.py:901] (1/2) Epoch 33, batch 2600, loss[loss=0.1375, simple_loss=0.227, pruned_loss=0.024, over 7222.00 frames. ], tot_loss[loss=0.1342, simple_loss=0.2151, pruned_loss=0.02669, over 1439000.54 frames. ], batch size: 99, lr: 4.97e-03, grad_scale: 8.0 +2023-03-21 07:03:02,772 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2565, 2.2897, 2.3429, 3.3828, 1.8384, 3.2816, 1.3182, 3.1275], + device='cuda:1'), covar=tensor([0.0215, 0.1347, 0.1675, 0.0198, 0.4051, 0.0249, 0.1212, 0.0517], + device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0253, 0.0265, 0.0204, 0.0254, 0.0211, 0.0233, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 07:03:05,795 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92975.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:03:08,629 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92981.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:03:12,109 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1584, 2.9445, 3.4613, 3.0759, 3.3193, 3.1641, 2.7156, 3.2290], + device='cuda:1'), covar=tensor([0.1316, 0.0677, 0.0726, 0.1691, 0.0708, 0.0958, 0.1880, 0.1284], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0064, 0.0049, 0.0048, 0.0048, 0.0048, 0.0065, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 07:03:14,107 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92992.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:03:20,320 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9956, 2.0994, 2.1941, 2.2377, 2.2247, 2.1117, 1.9989, 1.6525], + device='cuda:1'), covar=tensor([0.0521, 0.0670, 0.0343, 0.0197, 0.0515, 0.0523, 0.0319, 0.0395], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0035, 0.0037, 0.0034, 0.0034, 0.0033, 0.0036, 0.0037], + device='cuda:1'), out_proj_covar=tensor([9.1227e-05, 9.0126e-05, 9.1304e-05, 8.6357e-05, 8.7601e-05, 8.5726e-05, + 9.1532e-05, 9.3477e-05], device='cuda:1') +2023-03-21 07:03:23,310 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93010.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:03:27,662 INFO [train.py:901] (1/2) Epoch 33, batch 2650, loss[loss=0.1449, simple_loss=0.2246, pruned_loss=0.03256, over 7223.00 frames. ], tot_loss[loss=0.1345, simple_loss=0.2155, pruned_loss=0.02677, over 1442386.60 frames. ], batch size: 93, lr: 4.97e-03, grad_scale: 8.0 +2023-03-21 07:03:32,355 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0081, 3.9936, 3.4522, 3.6068, 3.1756, 2.3690, 2.0401, 4.1490], + device='cuda:1'), covar=tensor([0.0058, 0.0074, 0.0124, 0.0063, 0.0134, 0.0482, 0.0598, 0.0042], + device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0088, 0.0108, 0.0093, 0.0123, 0.0130, 0.0128, 0.0101], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 07:03:36,395 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:03:36,409 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:03:38,269 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.296e+02 1.709e+02 2.039e+02 2.473e+02 4.429e+02, threshold=4.078e+02, percent-clipped=1.0 +2023-03-21 07:03:42,405 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3863, 1.7342, 1.6202, 1.6883, 1.8299, 1.5656, 1.4945, 1.3461], + device='cuda:1'), covar=tensor([0.0211, 0.0170, 0.0197, 0.0123, 0.0124, 0.0181, 0.0165, 0.0183], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0032, 0.0031, 0.0033, 0.0031, 0.0030, 0.0034, 0.0041], + device='cuda:1'), out_proj_covar=tensor([3.9094e-05, 3.5369e-05, 3.5132e-05, 3.6304e-05, 3.4867e-05, 3.3410e-05, + 3.8184e-05, 4.5183e-05], device='cuda:1') +2023-03-21 07:03:52,586 INFO [train.py:901] (1/2) Epoch 33, batch 2700, loss[loss=0.1351, simple_loss=0.2117, pruned_loss=0.02926, over 7309.00 frames. ], tot_loss[loss=0.1344, simple_loss=0.2154, pruned_loss=0.02664, over 1442982.99 frames. ], batch size: 49, lr: 4.97e-03, grad_scale: 8.0 +2023-03-21 07:03:53,706 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93071.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:03:56,177 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93076.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:03:56,291 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7322, 3.1551, 3.6178, 3.4998, 3.1143, 2.8991, 3.6768, 2.7067], + device='cuda:1'), covar=tensor([0.0454, 0.0406, 0.0535, 0.0645, 0.0695, 0.0941, 0.0595, 0.2074], + device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0337, 0.0271, 0.0357, 0.0293, 0.0293, 0.0346, 0.0259], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:04:06,572 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93097.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 07:04:15,339 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93115.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:04:17,248 INFO [train.py:901] (1/2) Epoch 33, batch 2750, loss[loss=0.1369, simple_loss=0.2262, pruned_loss=0.02381, over 7315.00 frames. ], tot_loss[loss=0.1338, simple_loss=0.2149, pruned_loss=0.02632, over 1444099.41 frames. ], batch size: 59, lr: 4.97e-03, grad_scale: 8.0 +2023-03-21 07:04:27,762 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.702e+02 2.040e+02 2.360e+02 5.193e+02, threshold=4.081e+02, percent-clipped=1.0 +2023-03-21 07:04:34,348 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1526, 2.6427, 2.8511, 2.6232, 2.3547, 2.5113, 2.3706, 2.1926], + device='cuda:1'), covar=tensor([0.0643, 0.0563, 0.0279, 0.0156, 0.0864, 0.0522, 0.0287, 0.0275], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0034, 0.0036, 0.0033, 0.0033, 0.0032, 0.0036, 0.0036], + device='cuda:1'), out_proj_covar=tensor([8.9745e-05, 8.8328e-05, 8.9980e-05, 8.5103e-05, 8.6466e-05, 8.4696e-05, + 9.0570e-05, 9.1891e-05], device='cuda:1') +2023-03-21 07:04:41,615 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9282, 2.5304, 2.8560, 3.0672, 2.4472, 2.4814, 2.9073, 2.3193], + device='cuda:1'), covar=tensor([0.0524, 0.0518, 0.0674, 0.0631, 0.0668, 0.0907, 0.0696, 0.1897], + device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0337, 0.0272, 0.0358, 0.0294, 0.0292, 0.0346, 0.0258], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:04:41,913 INFO [train.py:901] (1/2) Epoch 33, batch 2800, loss[loss=0.1299, simple_loss=0.2152, pruned_loss=0.0223, over 7248.00 frames. ], tot_loss[loss=0.1344, simple_loss=0.2155, pruned_loss=0.02669, over 1444693.27 frames. ], batch size: 89, lr: 4.97e-03, grad_scale: 8.0 +2023-03-21 07:04:45,825 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93176.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:04:51,033 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93187.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:05:07,808 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 07:05:07,873 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 07:05:07,898 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 07:05:07,919 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 07:05:08,089 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 07:05:08,109 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 07:05:08,119 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 07:05:08,221 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 07:05:08,297 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 07:05:08,312 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 07:05:08,327 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 07:05:08,809 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 07:05:08,814 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 07:05:08,821 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 07:05:08,973 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 07:05:08,976 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 07:05:09,033 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 07:05:14,375 INFO [train.py:901] (1/2) Epoch 34, batch 0, loss[loss=0.1534, simple_loss=0.2289, pruned_loss=0.03891, over 7314.00 frames. ], tot_loss[loss=0.1534, simple_loss=0.2289, pruned_loss=0.03891, over 7314.00 frames. ], batch size: 83, lr: 4.89e-03, grad_scale: 8.0 +2023-03-21 07:05:14,375 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 07:05:21,627 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0416, 4.6327, 4.2839, 4.9774, 4.7644, 5.0311, 4.7705, 4.8348], + device='cuda:1'), covar=tensor([0.0519, 0.1977, 0.1964, 0.1049, 0.0832, 0.0820, 0.0507, 0.0727], + device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0381, 0.0290, 0.0299, 0.0222, 0.0358, 0.0218, 0.0267], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:05:25,267 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6893, 2.9371, 2.5855, 3.9041, 1.9592, 3.5726, 1.8413, 3.2738], + device='cuda:1'), covar=tensor([0.0176, 0.0984, 0.1706, 0.0145, 0.4077, 0.0187, 0.1106, 0.0405], + device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0252, 0.0263, 0.0204, 0.0252, 0.0210, 0.0231, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 07:05:40,677 INFO [train.py:935] (1/2) Epoch 34, validation: loss=0.1638, simple_loss=0.2553, pruned_loss=0.03615, over 1622729.00 frames. +2023-03-21 07:05:40,679 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 07:05:49,819 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 07:05:51,348 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2374, 4.1791, 4.0463, 3.8562, 3.5208, 2.4400, 2.1275, 4.3853], + device='cuda:1'), covar=tensor([0.0082, 0.0104, 0.0098, 0.0080, 0.0143, 0.0601, 0.0642, 0.0063], + device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0089, 0.0110, 0.0094, 0.0125, 0.0132, 0.0129, 0.0102], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 07:05:59,707 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 07:06:05,298 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.686e+02 1.995e+02 2.522e+02 6.389e+02, threshold=3.989e+02, percent-clipped=1.0 +2023-03-21 07:06:06,307 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 07:06:06,788 INFO [train.py:901] (1/2) Epoch 34, batch 50, loss[loss=0.1351, simple_loss=0.2195, pruned_loss=0.02531, over 7286.00 frames. ], tot_loss[loss=0.1331, simple_loss=0.2145, pruned_loss=0.02587, over 327098.52 frames. ], batch size: 80, lr: 4.89e-03, grad_scale: 8.0 +2023-03-21 07:06:08,830 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 07:06:10,018 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8229, 1.7279, 2.0741, 2.5169, 2.1257, 2.2962, 2.0898, 2.3590], + device='cuda:1'), covar=tensor([0.2960, 0.3666, 0.3166, 0.1058, 0.2247, 0.7114, 0.2050, 0.4433], + device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0075, 0.0063, 0.0059, 0.0059, 0.0059, 0.0099, 0.0061], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:06:11,397 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 07:06:13,009 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93255.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:06:16,217 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.18 vs. limit=5.0 +2023-03-21 07:06:25,916 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93281.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:06:29,511 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93287.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:06:29,957 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 07:06:30,485 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 07:06:33,138 INFO [train.py:901] (1/2) Epoch 34, batch 100, loss[loss=0.1458, simple_loss=0.2282, pruned_loss=0.03167, over 7298.00 frames. ], tot_loss[loss=0.1328, simple_loss=0.2138, pruned_loss=0.02589, over 575515.53 frames. ], batch size: 77, lr: 4.89e-03, grad_scale: 8.0 +2023-03-21 07:06:51,315 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=93329.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:06:52,372 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93331.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:06:56,804 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.175e+02 1.737e+02 2.034e+02 2.402e+02 4.556e+02, threshold=4.067e+02, percent-clipped=1.0 +2023-03-21 07:06:58,270 INFO [train.py:901] (1/2) Epoch 34, batch 150, loss[loss=0.1782, simple_loss=0.2585, pruned_loss=0.04897, over 6750.00 frames. ], tot_loss[loss=0.1331, simple_loss=0.2144, pruned_loss=0.02587, over 768640.67 frames. ], batch size: 106, lr: 4.89e-03, grad_scale: 8.0 +2023-03-21 07:07:09,971 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93366.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:07:16,255 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93376.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:07:24,197 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93392.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 07:07:24,604 INFO [train.py:901] (1/2) Epoch 34, batch 200, loss[loss=0.1478, simple_loss=0.2274, pruned_loss=0.03407, over 7285.00 frames. ], tot_loss[loss=0.1324, simple_loss=0.2133, pruned_loss=0.02573, over 919232.22 frames. ], batch size: 57, lr: 4.89e-03, grad_scale: 8.0 +2023-03-21 07:07:29,099 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 07:07:33,776 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 07:07:39,298 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 07:07:40,835 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=93424.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:07:48,750 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.225e+02 1.705e+02 1.990e+02 2.336e+02 5.774e+02, threshold=3.979e+02, percent-clipped=1.0 +2023-03-21 07:07:49,356 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6121, 4.2037, 4.0448, 4.5790, 4.4564, 4.5569, 4.1236, 4.1505], + device='cuda:1'), covar=tensor([0.0916, 0.2478, 0.1945, 0.1137, 0.0760, 0.1135, 0.0696, 0.1326], + device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0378, 0.0287, 0.0297, 0.0219, 0.0356, 0.0216, 0.0265], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:07:50,290 INFO [train.py:901] (1/2) Epoch 34, batch 250, loss[loss=0.1219, simple_loss=0.1927, pruned_loss=0.02554, over 7067.00 frames. ], tot_loss[loss=0.1332, simple_loss=0.2141, pruned_loss=0.02619, over 1035293.73 frames. ], batch size: 35, lr: 4.89e-03, grad_scale: 8.0 +2023-03-21 07:07:52,282 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 07:08:05,525 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93471.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:08:12,977 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 07:08:13,574 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93487.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:08:16,419 INFO [train.py:901] (1/2) Epoch 34, batch 300, loss[loss=0.1676, simple_loss=0.241, pruned_loss=0.04709, over 7343.00 frames. ], tot_loss[loss=0.1344, simple_loss=0.2151, pruned_loss=0.02687, over 1128396.51 frames. ], batch size: 54, lr: 4.89e-03, grad_scale: 8.0 +2023-03-21 07:08:21,973 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 07:08:37,529 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=93535.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:08:39,925 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.325e+02 2.018e+02 2.343e+02 2.755e+02 4.268e+02, threshold=4.687e+02, percent-clipped=3.0 +2023-03-21 07:08:42,066 INFO [train.py:901] (1/2) Epoch 34, batch 350, loss[loss=0.1486, simple_loss=0.2242, pruned_loss=0.03655, over 7269.00 frames. ], tot_loss[loss=0.1337, simple_loss=0.2146, pruned_loss=0.02645, over 1198656.08 frames. ], batch size: 52, lr: 4.89e-03, grad_scale: 8.0 +2023-03-21 07:08:48,955 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93555.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:08:54,985 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7305, 1.4199, 1.9775, 2.3839, 1.9453, 2.1604, 1.7825, 2.2073], + device='cuda:1'), covar=tensor([0.2823, 0.5549, 0.1660, 0.2146, 0.1512, 0.2592, 0.2211, 0.4352], + device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0076, 0.0065, 0.0060, 0.0059, 0.0059, 0.0099, 0.0062], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:08:57,341 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 07:09:04,882 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93587.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:09:07,746 INFO [train.py:901] (1/2) Epoch 34, batch 400, loss[loss=0.123, simple_loss=0.207, pruned_loss=0.01947, over 7294.00 frames. ], tot_loss[loss=0.1346, simple_loss=0.2153, pruned_loss=0.0269, over 1253953.89 frames. ], batch size: 57, lr: 4.88e-03, grad_scale: 8.0 +2023-03-21 07:09:13,157 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=93603.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:09:28,705 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93631.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:09:30,700 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=93635.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:09:33,108 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.211e+02 1.661e+02 1.987e+02 2.340e+02 3.942e+02, threshold=3.975e+02, percent-clipped=0.0 +2023-03-21 07:09:34,667 INFO [train.py:901] (1/2) Epoch 34, batch 450, loss[loss=0.1304, simple_loss=0.2178, pruned_loss=0.02154, over 7290.00 frames. ], tot_loss[loss=0.134, simple_loss=0.2147, pruned_loss=0.02659, over 1294299.90 frames. ], batch size: 68, lr: 4.88e-03, grad_scale: 8.0 +2023-03-21 07:09:39,720 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 07:09:40,224 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 07:09:46,321 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93666.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:09:52,799 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=93679.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:09:59,401 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93692.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:09:59,821 INFO [train.py:901] (1/2) Epoch 34, batch 500, loss[loss=0.1053, simple_loss=0.1697, pruned_loss=0.02042, over 6426.00 frames. ], tot_loss[loss=0.134, simple_loss=0.215, pruned_loss=0.02656, over 1329841.69 frames. ], batch size: 28, lr: 4.88e-03, grad_scale: 8.0 +2023-03-21 07:10:10,391 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=93714.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:10:11,994 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 07:10:13,544 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 07:10:14,695 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 07:10:16,722 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 07:10:21,190 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 07:10:24,669 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.204e+02 1.702e+02 2.032e+02 2.310e+02 3.971e+02, threshold=4.063e+02, percent-clipped=0.0 +2023-03-21 07:10:24,757 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=93740.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:10:26,219 INFO [train.py:901] (1/2) Epoch 34, batch 550, loss[loss=0.1457, simple_loss=0.2265, pruned_loss=0.03247, over 7344.00 frames. ], tot_loss[loss=0.1336, simple_loss=0.2143, pruned_loss=0.0264, over 1354914.42 frames. ], batch size: 54, lr: 4.88e-03, grad_scale: 8.0 +2023-03-21 07:10:32,131 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 07:10:39,753 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93770.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:10:40,177 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 07:10:40,272 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93771.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:10:43,637 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 07:10:50,630 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 07:10:51,078 INFO [train.py:901] (1/2) Epoch 34, batch 600, loss[loss=0.1545, simple_loss=0.2342, pruned_loss=0.03741, over 7325.00 frames. ], tot_loss[loss=0.1333, simple_loss=0.2141, pruned_loss=0.02624, over 1374991.85 frames. ], batch size: 83, lr: 4.88e-03, grad_scale: 8.0 +2023-03-21 07:11:05,805 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=93819.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:11:08,882 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 07:11:12,154 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93831.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:11:16,501 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.397e+02 1.841e+02 2.134e+02 2.382e+02 3.944e+02, threshold=4.268e+02, percent-clipped=0.0 +2023-03-21 07:11:18,115 INFO [train.py:901] (1/2) Epoch 34, batch 650, loss[loss=0.1352, simple_loss=0.2114, pruned_loss=0.02954, over 7310.00 frames. ], tot_loss[loss=0.1331, simple_loss=0.2137, pruned_loss=0.02628, over 1388355.87 frames. ], batch size: 49, lr: 4.88e-03, grad_scale: 8.0 +2023-03-21 07:11:18,130 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 07:11:21,628 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 +2023-03-21 07:11:34,677 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 07:11:43,901 INFO [train.py:901] (1/2) Epoch 34, batch 700, loss[loss=0.134, simple_loss=0.2203, pruned_loss=0.02388, over 7336.00 frames. ], tot_loss[loss=0.1336, simple_loss=0.2144, pruned_loss=0.02647, over 1400873.85 frames. ], batch size: 51, lr: 4.88e-03, grad_scale: 8.0 +2023-03-21 07:11:44,509 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 07:11:47,679 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1155, 3.5787, 4.0829, 4.1233, 4.2268, 4.1219, 4.3726, 4.1155], + device='cuda:1'), covar=tensor([0.0025, 0.0097, 0.0033, 0.0027, 0.0025, 0.0030, 0.0024, 0.0053], + device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0067, 0.0056, 0.0054, 0.0054, 0.0059, 0.0048, 0.0074], + device='cuda:1'), out_proj_covar=tensor([8.0814e-05, 1.4049e-04, 1.0538e-04, 9.6002e-05, 9.4684e-05, 1.0587e-04, + 9.4610e-05, 1.4103e-04], device='cuda:1') +2023-03-21 07:12:08,218 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.210e+02 1.694e+02 2.006e+02 2.347e+02 4.324e+02, threshold=4.012e+02, percent-clipped=1.0 +2023-03-21 07:12:08,242 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 07:12:08,758 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 07:12:09,766 INFO [train.py:901] (1/2) Epoch 34, batch 750, loss[loss=0.1362, simple_loss=0.2228, pruned_loss=0.02481, over 7238.00 frames. ], tot_loss[loss=0.1337, simple_loss=0.2147, pruned_loss=0.02638, over 1411011.78 frames. ], batch size: 93, lr: 4.88e-03, grad_scale: 8.0 +2023-03-21 07:12:22,899 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 07:12:27,670 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 07:12:30,457 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93981.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:12:34,821 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 07:12:35,813 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 07:12:36,284 INFO [train.py:901] (1/2) Epoch 34, batch 800, loss[loss=0.1351, simple_loss=0.2184, pruned_loss=0.02593, over 7284.00 frames. ], tot_loss[loss=0.1334, simple_loss=0.2145, pruned_loss=0.02621, over 1415445.23 frames. ], batch size: 68, lr: 4.87e-03, grad_scale: 8.0 +2023-03-21 07:12:38,504 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6427, 3.1164, 3.6261, 3.5225, 3.1368, 3.0983, 3.6843, 2.8154], + device='cuda:1'), covar=tensor([0.0316, 0.0423, 0.0543, 0.0565, 0.0615, 0.0785, 0.0451, 0.1842], + device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0333, 0.0269, 0.0355, 0.0290, 0.0288, 0.0342, 0.0254], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:12:45,452 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5927, 3.8127, 3.6658, 3.7926, 3.4748, 3.7849, 4.0464, 4.1075], + device='cuda:1'), covar=tensor([0.0275, 0.0185, 0.0234, 0.0197, 0.0461, 0.0319, 0.0296, 0.0207], + device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0121, 0.0113, 0.0116, 0.0109, 0.0100, 0.0096, 0.0095], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:12:46,859 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 07:13:00,410 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.299e+02 1.747e+02 2.063e+02 2.429e+02 4.246e+02, threshold=4.125e+02, percent-clipped=1.0 +2023-03-21 07:13:01,549 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94042.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:13:01,927 INFO [train.py:901] (1/2) Epoch 34, batch 850, loss[loss=0.1332, simple_loss=0.2215, pruned_loss=0.02246, over 7262.00 frames. ], tot_loss[loss=0.1333, simple_loss=0.2143, pruned_loss=0.02619, over 1420044.60 frames. ], batch size: 55, lr: 4.87e-03, grad_scale: 8.0 +2023-03-21 07:13:05,497 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 07:13:05,999 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 07:13:10,557 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 07:13:15,545 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 +2023-03-21 07:13:15,751 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 07:13:16,113 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-21 07:13:28,326 INFO [train.py:901] (1/2) Epoch 34, batch 900, loss[loss=0.1182, simple_loss=0.2043, pruned_loss=0.016, over 7308.00 frames. ], tot_loss[loss=0.1328, simple_loss=0.2136, pruned_loss=0.02598, over 1424196.72 frames. ], batch size: 75, lr: 4.87e-03, grad_scale: 8.0 +2023-03-21 07:13:45,073 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94126.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:13:52,191 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.701e+02 1.978e+02 2.381e+02 4.265e+02, threshold=3.957e+02, percent-clipped=2.0 +2023-03-21 07:13:53,732 INFO [train.py:901] (1/2) Epoch 34, batch 950, loss[loss=0.1325, simple_loss=0.2126, pruned_loss=0.02617, over 7353.00 frames. ], tot_loss[loss=0.1331, simple_loss=0.214, pruned_loss=0.02612, over 1428538.27 frames. ], batch size: 73, lr: 4.87e-03, grad_scale: 8.0 +2023-03-21 07:13:54,246 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 07:14:04,089 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3853, 2.4166, 2.2578, 3.5170, 1.7447, 3.3737, 1.4109, 3.2726], + device='cuda:1'), covar=tensor([0.0189, 0.1381, 0.1864, 0.0196, 0.3952, 0.0204, 0.1243, 0.0402], + device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0247, 0.0263, 0.0203, 0.0250, 0.0206, 0.0230, 0.0223], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 07:14:07,341 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-03-21 07:14:14,579 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5401, 2.8797, 2.4228, 2.6766, 2.8120, 2.4400, 2.8024, 2.7662], + device='cuda:1'), covar=tensor([0.0772, 0.0494, 0.0788, 0.0775, 0.0924, 0.0592, 0.0884, 0.0693], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0056, 0.0064, 0.0056, 0.0053, 0.0058, 0.0054, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:14:18,010 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 07:14:20,043 INFO [train.py:901] (1/2) Epoch 34, batch 1000, loss[loss=0.1502, simple_loss=0.2274, pruned_loss=0.03647, over 7298.00 frames. ], tot_loss[loss=0.1332, simple_loss=0.2141, pruned_loss=0.02608, over 1431900.73 frames. ], batch size: 86, lr: 4.87e-03, grad_scale: 8.0 +2023-03-21 07:14:35,668 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94223.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 07:14:38,133 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 07:14:44,828 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.337e+02 1.717e+02 1.958e+02 2.370e+02 4.573e+02, threshold=3.915e+02, percent-clipped=2.0 +2023-03-21 07:14:46,365 INFO [train.py:901] (1/2) Epoch 34, batch 1050, loss[loss=0.1046, simple_loss=0.1657, pruned_loss=0.0218, over 5752.00 frames. ], tot_loss[loss=0.1326, simple_loss=0.2134, pruned_loss=0.02588, over 1430064.64 frames. ], batch size: 25, lr: 4.87e-03, grad_scale: 8.0 +2023-03-21 07:15:00,007 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 07:15:04,467 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 07:15:07,682 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94284.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 07:15:12,024 INFO [train.py:901] (1/2) Epoch 34, batch 1100, loss[loss=0.1314, simple_loss=0.2184, pruned_loss=0.02222, over 7227.00 frames. ], tot_loss[loss=0.1331, simple_loss=0.2142, pruned_loss=0.02602, over 1432676.08 frames. ], batch size: 93, lr: 4.87e-03, grad_scale: 8.0 +2023-03-21 07:15:32,048 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 07:15:32,582 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 07:15:35,618 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94337.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:15:37,014 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.720e+02 1.980e+02 2.346e+02 4.157e+02, threshold=3.960e+02, percent-clipped=1.0 +2023-03-21 07:15:38,529 INFO [train.py:901] (1/2) Epoch 34, batch 1150, loss[loss=0.1418, simple_loss=0.2176, pruned_loss=0.03302, over 7297.00 frames. ], tot_loss[loss=0.1337, simple_loss=0.2148, pruned_loss=0.02632, over 1436740.62 frames. ], batch size: 68, lr: 4.87e-03, grad_scale: 8.0 +2023-03-21 07:15:41,094 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94348.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:15:44,474 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 07:15:44,971 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 07:16:03,413 INFO [train.py:901] (1/2) Epoch 34, batch 1200, loss[loss=0.1357, simple_loss=0.2213, pruned_loss=0.02511, over 7314.00 frames. ], tot_loss[loss=0.1341, simple_loss=0.215, pruned_loss=0.02661, over 1436160.63 frames. ], batch size: 83, lr: 4.86e-03, grad_scale: 8.0 +2023-03-21 07:16:12,567 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94409.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 07:16:18,514 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 07:16:21,209 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-03-21 07:16:21,571 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94426.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:16:28,467 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.293e+02 1.769e+02 2.087e+02 2.412e+02 4.266e+02, threshold=4.175e+02, percent-clipped=1.0 +2023-03-21 07:16:29,983 INFO [train.py:901] (1/2) Epoch 34, batch 1250, loss[loss=0.1397, simple_loss=0.2223, pruned_loss=0.02858, over 7311.00 frames. ], tot_loss[loss=0.1341, simple_loss=0.215, pruned_loss=0.02662, over 1438121.05 frames. ], batch size: 59, lr: 4.86e-03, grad_scale: 8.0 +2023-03-21 07:16:30,386 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 +2023-03-21 07:16:42,508 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 07:16:45,598 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=94474.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:16:47,100 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 07:16:48,087 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 07:16:55,089 INFO [train.py:901] (1/2) Epoch 34, batch 1300, loss[loss=0.1319, simple_loss=0.2178, pruned_loss=0.02304, over 7300.00 frames. ], tot_loss[loss=0.1344, simple_loss=0.2153, pruned_loss=0.02674, over 1439705.93 frames. ], batch size: 77, lr: 4.86e-03, grad_scale: 8.0 +2023-03-21 07:17:07,657 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-03-21 07:17:12,401 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 07:17:14,903 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 07:17:18,399 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 07:17:19,913 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.017e+02 1.844e+02 2.157e+02 2.540e+02 4.184e+02, threshold=4.314e+02, percent-clipped=1.0 +2023-03-21 07:17:21,507 INFO [train.py:901] (1/2) Epoch 34, batch 1350, loss[loss=0.1344, simple_loss=0.2175, pruned_loss=0.02566, over 7247.00 frames. ], tot_loss[loss=0.134, simple_loss=0.2147, pruned_loss=0.02661, over 1439979.52 frames. ], batch size: 55, lr: 4.86e-03, grad_scale: 8.0 +2023-03-21 07:17:26,288 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-03-21 07:17:28,049 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 07:17:39,611 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94579.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 07:17:47,201 INFO [train.py:901] (1/2) Epoch 34, batch 1400, loss[loss=0.1231, simple_loss=0.2077, pruned_loss=0.01923, over 7322.00 frames. ], tot_loss[loss=0.1343, simple_loss=0.2154, pruned_loss=0.02663, over 1442664.74 frames. ], batch size: 59, lr: 4.86e-03, grad_scale: 8.0 +2023-03-21 07:18:01,606 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 07:18:10,200 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94637.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:18:11,544 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.216e+02 1.753e+02 2.058e+02 2.553e+02 5.277e+02, threshold=4.117e+02, percent-clipped=1.0 +2023-03-21 07:18:13,066 INFO [train.py:901] (1/2) Epoch 34, batch 1450, loss[loss=0.145, simple_loss=0.2227, pruned_loss=0.03362, over 7311.00 frames. ], tot_loss[loss=0.1344, simple_loss=0.2153, pruned_loss=0.02677, over 1442211.74 frames. ], batch size: 59, lr: 4.86e-03, grad_scale: 8.0 +2023-03-21 07:18:24,509 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 07:18:35,358 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=94685.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:18:39,195 INFO [train.py:901] (1/2) Epoch 34, batch 1500, loss[loss=0.1467, simple_loss=0.2341, pruned_loss=0.02961, over 6719.00 frames. ], tot_loss[loss=0.1347, simple_loss=0.2156, pruned_loss=0.02692, over 1440227.13 frames. ], batch size: 106, lr: 4.86e-03, grad_scale: 8.0 +2023-03-21 07:18:41,788 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 07:18:43,863 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9063, 4.3218, 4.4234, 4.3612, 4.3719, 3.9793, 4.4118, 4.3527], + device='cuda:1'), covar=tensor([0.0520, 0.0464, 0.0356, 0.0490, 0.0310, 0.0440, 0.0360, 0.0402], + device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0248, 0.0190, 0.0189, 0.0149, 0.0221, 0.0196, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:18:44,909 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94704.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 07:19:02,124 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-03-21 07:19:02,837 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.786e+02 2.193e+02 2.482e+02 4.700e+02, threshold=4.385e+02, percent-clipped=1.0 +2023-03-21 07:19:04,324 INFO [train.py:901] (1/2) Epoch 34, batch 1550, loss[loss=0.1448, simple_loss=0.2249, pruned_loss=0.03232, over 7299.00 frames. ], tot_loss[loss=0.1352, simple_loss=0.216, pruned_loss=0.02716, over 1440122.38 frames. ], batch size: 68, lr: 4.85e-03, grad_scale: 8.0 +2023-03-21 07:19:05,377 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 07:19:11,074 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0209, 3.7491, 3.7168, 3.7996, 3.1587, 3.6563, 3.8737, 3.4373], + device='cuda:1'), covar=tensor([0.0213, 0.0269, 0.0202, 0.0253, 0.0892, 0.0202, 0.0288, 0.0269], + device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0099, 0.0098, 0.0087, 0.0173, 0.0103, 0.0102, 0.0108], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:19:30,392 INFO [train.py:901] (1/2) Epoch 34, batch 1600, loss[loss=0.1371, simple_loss=0.225, pruned_loss=0.02457, over 7344.00 frames. ], tot_loss[loss=0.1353, simple_loss=0.2159, pruned_loss=0.02732, over 1441582.17 frames. ], batch size: 54, lr: 4.85e-03, grad_scale: 8.0 +2023-03-21 07:19:37,382 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 07:19:38,376 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 07:19:41,399 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 07:19:50,969 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 07:19:51,612 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4336, 3.4690, 2.4842, 3.9785, 3.1169, 3.5142, 1.7225, 2.4465], + device='cuda:1'), covar=tensor([0.0427, 0.0669, 0.2400, 0.0558, 0.0433, 0.0647, 0.3595, 0.1846], + device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0256, 0.0283, 0.0272, 0.0271, 0.0265, 0.0240, 0.0265], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 07:19:54,406 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 1.787e+02 1.989e+02 2.299e+02 4.264e+02, threshold=3.978e+02, percent-clipped=0.0 +2023-03-21 07:19:54,468 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 07:19:56,602 INFO [train.py:901] (1/2) Epoch 34, batch 1650, loss[loss=0.1256, simple_loss=0.2071, pruned_loss=0.02202, over 7339.00 frames. ], tot_loss[loss=0.1351, simple_loss=0.2158, pruned_loss=0.02722, over 1441290.62 frames. ], batch size: 54, lr: 4.85e-03, grad_scale: 8.0 +2023-03-21 07:19:57,399 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-03-21 07:20:03,207 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 07:20:03,745 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.5493, 5.0795, 4.9462, 5.5211, 5.3071, 5.4908, 4.8801, 5.1242], + device='cuda:1'), covar=tensor([0.0564, 0.1976, 0.1695, 0.0889, 0.0765, 0.0973, 0.0698, 0.1065], + device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0383, 0.0289, 0.0297, 0.0223, 0.0358, 0.0219, 0.0266], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:20:15,549 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94879.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 07:20:20,960 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 07:20:22,474 INFO [train.py:901] (1/2) Epoch 34, batch 1700, loss[loss=0.1343, simple_loss=0.2159, pruned_loss=0.02637, over 7314.00 frames. ], tot_loss[loss=0.1341, simple_loss=0.2147, pruned_loss=0.0268, over 1438695.83 frames. ], batch size: 80, lr: 4.85e-03, grad_scale: 16.0 +2023-03-21 07:20:25,068 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 07:20:35,530 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 07:20:40,284 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=94927.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 07:20:46,782 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.214e+02 1.668e+02 1.961e+02 2.375e+02 4.314e+02, threshold=3.923e+02, percent-clipped=2.0 +2023-03-21 07:20:47,440 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0176, 2.6642, 3.2536, 2.9063, 3.2121, 2.9330, 2.4458, 3.0435], + device='cuda:1'), covar=tensor([0.1580, 0.0984, 0.0934, 0.1628, 0.0757, 0.1135, 0.2362, 0.1143], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0065, 0.0049, 0.0049, 0.0047, 0.0049, 0.0066, 0.0049], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 07:20:48,256 INFO [train.py:901] (1/2) Epoch 34, batch 1750, loss[loss=0.1465, simple_loss=0.2138, pruned_loss=0.03961, over 7241.00 frames. ], tot_loss[loss=0.134, simple_loss=0.2146, pruned_loss=0.02671, over 1441160.74 frames. ], batch size: 55, lr: 4.85e-03, grad_scale: 16.0 +2023-03-21 07:20:54,242 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-03-21 07:21:01,608 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 07:21:02,567 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 07:21:13,875 INFO [train.py:901] (1/2) Epoch 34, batch 1800, loss[loss=0.129, simple_loss=0.215, pruned_loss=0.0215, over 7288.00 frames. ], tot_loss[loss=0.1339, simple_loss=0.2145, pruned_loss=0.02662, over 1439538.04 frames. ], batch size: 68, lr: 4.85e-03, grad_scale: 8.0 +2023-03-21 07:21:19,854 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95004.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:21:24,818 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 07:21:39,040 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 07:21:39,510 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.226e+02 1.793e+02 2.137e+02 2.599e+02 4.200e+02, threshold=4.274e+02, percent-clipped=3.0 +2023-03-21 07:21:40,601 INFO [train.py:901] (1/2) Epoch 34, batch 1850, loss[loss=0.1043, simple_loss=0.1829, pruned_loss=0.01282, over 7196.00 frames. ], tot_loss[loss=0.1344, simple_loss=0.2147, pruned_loss=0.02709, over 1436334.28 frames. ], batch size: 39, lr: 4.85e-03, grad_scale: 8.0 +2023-03-21 07:21:45,127 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=95052.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:21:48,027 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 07:21:54,847 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-03-21 07:22:04,946 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 07:22:05,418 INFO [train.py:901] (1/2) Epoch 34, batch 1900, loss[loss=0.1308, simple_loss=0.2097, pruned_loss=0.02595, over 7248.00 frames. ], tot_loss[loss=0.1346, simple_loss=0.2148, pruned_loss=0.02722, over 1435424.16 frames. ], batch size: 55, lr: 4.85e-03, grad_scale: 8.0 +2023-03-21 07:22:05,587 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8380, 3.2747, 2.6496, 3.0147, 3.0423, 2.7400, 3.2094, 3.0133], + device='cuda:1'), covar=tensor([0.0871, 0.0728, 0.1148, 0.1138, 0.1146, 0.0720, 0.1004, 0.0818], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0057, 0.0066, 0.0058, 0.0055, 0.0060, 0.0055, 0.0054], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:22:10,699 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1879, 3.8439, 3.8029, 3.9452, 3.7993, 3.8061, 4.1037, 3.5342], + device='cuda:1'), covar=tensor([0.0131, 0.0183, 0.0142, 0.0165, 0.0537, 0.0133, 0.0168, 0.0239], + device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0101, 0.0100, 0.0088, 0.0176, 0.0106, 0.0104, 0.0110], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:22:12,136 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95105.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 07:22:30,128 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 07:22:30,536 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.839e+02 2.066e+02 2.358e+02 4.845e+02, threshold=4.133e+02, percent-clipped=1.0 +2023-03-21 07:22:31,617 INFO [train.py:901] (1/2) Epoch 34, batch 1950, loss[loss=0.1044, simple_loss=0.1844, pruned_loss=0.01217, over 7190.00 frames. ], tot_loss[loss=0.1343, simple_loss=0.2146, pruned_loss=0.02699, over 1436745.57 frames. ], batch size: 39, lr: 4.84e-03, grad_scale: 8.0 +2023-03-21 07:22:41,116 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 07:22:41,389 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-03-21 07:22:43,219 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95166.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 07:22:46,058 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 07:22:46,575 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 07:22:57,218 INFO [train.py:901] (1/2) Epoch 34, batch 2000, loss[loss=0.1348, simple_loss=0.2099, pruned_loss=0.02988, over 7318.00 frames. ], tot_loss[loss=0.1345, simple_loss=0.2152, pruned_loss=0.02691, over 1439625.64 frames. ], batch size: 49, lr: 4.84e-03, grad_scale: 8.0 +2023-03-21 07:23:00,432 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4402, 1.6415, 1.4491, 1.5715, 1.6205, 1.5804, 1.5960, 1.2463], + device='cuda:1'), covar=tensor([0.0158, 0.0126, 0.0293, 0.0135, 0.0132, 0.0114, 0.0143, 0.0190], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0033, 0.0032, 0.0034, 0.0032, 0.0031, 0.0034, 0.0040], + device='cuda:1'), out_proj_covar=tensor([3.9217e-05, 3.6469e-05, 3.6464e-05, 3.7402e-05, 3.5926e-05, 3.4634e-05, + 3.8952e-05, 4.4947e-05], device='cuda:1') +2023-03-21 07:23:03,149 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 07:23:14,838 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 07:23:15,225 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 +2023-03-21 07:23:21,688 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 +2023-03-21 07:23:22,278 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.193e+02 1.722e+02 2.097e+02 2.467e+02 4.921e+02, threshold=4.194e+02, percent-clipped=3.0 +2023-03-21 07:23:22,820 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 07:23:23,285 INFO [train.py:901] (1/2) Epoch 34, batch 2050, loss[loss=0.1482, simple_loss=0.2221, pruned_loss=0.03718, over 7262.00 frames. ], tot_loss[loss=0.1339, simple_loss=0.2146, pruned_loss=0.02663, over 1439664.20 frames. ], batch size: 47, lr: 4.84e-03, grad_scale: 8.0 +2023-03-21 07:23:26,598 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-03-21 07:23:32,026 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95260.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:23:36,076 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95268.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:23:45,025 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-03-21 07:23:49,093 INFO [train.py:901] (1/2) Epoch 34, batch 2100, loss[loss=0.1307, simple_loss=0.2185, pruned_loss=0.02141, over 7294.00 frames. ], tot_loss[loss=0.134, simple_loss=0.2148, pruned_loss=0.02659, over 1439562.63 frames. ], batch size: 80, lr: 4.84e-03, grad_scale: 8.0 +2023-03-21 07:23:53,426 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.13 vs. limit=5.0 +2023-03-21 07:23:56,733 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 07:23:59,735 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 07:24:04,007 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95321.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:24:08,030 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95329.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:24:12,964 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95339.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:24:13,820 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.277e+02 1.746e+02 2.026e+02 2.348e+02 4.750e+02, threshold=4.051e+02, percent-clipped=2.0 +2023-03-21 07:24:14,890 INFO [train.py:901] (1/2) Epoch 34, batch 2150, loss[loss=0.1062, simple_loss=0.1827, pruned_loss=0.01482, over 6975.00 frames. ], tot_loss[loss=0.134, simple_loss=0.2146, pruned_loss=0.02664, over 1438091.54 frames. ], batch size: 35, lr: 4.84e-03, grad_scale: 8.0 +2023-03-21 07:24:41,432 INFO [train.py:901] (1/2) Epoch 34, batch 2200, loss[loss=0.1413, simple_loss=0.2252, pruned_loss=0.02876, over 7266.00 frames. ], tot_loss[loss=0.1344, simple_loss=0.2151, pruned_loss=0.02692, over 1440057.98 frames. ], batch size: 70, lr: 4.84e-03, grad_scale: 8.0 +2023-03-21 07:24:45,435 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95400.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:24:45,851 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 07:25:00,898 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6678, 3.4001, 3.3398, 3.4184, 2.9797, 2.9926, 3.7674, 2.4861], + device='cuda:1'), covar=tensor([0.0452, 0.0713, 0.0638, 0.0658, 0.0893, 0.1081, 0.0610, 0.2531], + device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0339, 0.0275, 0.0358, 0.0294, 0.0290, 0.0349, 0.0257], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:25:02,999 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 +2023-03-21 07:25:05,777 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.237e+02 1.719e+02 1.925e+02 2.336e+02 3.673e+02, threshold=3.850e+02, percent-clipped=0.0 +2023-03-21 07:25:06,818 INFO [train.py:901] (1/2) Epoch 34, batch 2250, loss[loss=0.1502, simple_loss=0.2264, pruned_loss=0.03702, over 7302.00 frames. ], tot_loss[loss=0.1339, simple_loss=0.2145, pruned_loss=0.02665, over 1439117.36 frames. ], batch size: 49, lr: 4.84e-03, grad_scale: 8.0 +2023-03-21 07:25:11,241 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 +2023-03-21 07:25:16,511 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95461.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 07:25:20,922 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 07:25:20,933 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 07:25:32,905 INFO [train.py:901] (1/2) Epoch 34, batch 2300, loss[loss=0.126, simple_loss=0.2099, pruned_loss=0.021, over 7226.00 frames. ], tot_loss[loss=0.1339, simple_loss=0.2146, pruned_loss=0.0266, over 1441237.83 frames. ], batch size: 45, lr: 4.84e-03, grad_scale: 8.0 +2023-03-21 07:25:33,863 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 07:25:41,874 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5155, 4.0692, 4.2067, 4.3458, 4.2077, 4.2420, 4.5329, 4.0055], + device='cuda:1'), covar=tensor([0.0218, 0.0180, 0.0134, 0.0140, 0.0453, 0.0102, 0.0137, 0.0165], + device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0100, 0.0099, 0.0087, 0.0173, 0.0106, 0.0104, 0.0109], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:25:47,388 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8229, 1.5544, 2.0384, 2.4237, 2.0753, 2.2154, 1.9790, 2.2857], + device='cuda:1'), covar=tensor([0.3143, 0.3580, 0.1955, 0.1356, 0.2011, 0.3855, 0.1816, 0.1838], + device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0076, 0.0066, 0.0060, 0.0059, 0.0060, 0.0101, 0.0062], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:25:57,305 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.302e+02 1.783e+02 2.106e+02 2.494e+02 4.244e+02, threshold=4.213e+02, percent-clipped=2.0 +2023-03-21 07:25:58,315 INFO [train.py:901] (1/2) Epoch 34, batch 2350, loss[loss=0.1362, simple_loss=0.2191, pruned_loss=0.02663, over 7260.00 frames. ], tot_loss[loss=0.1344, simple_loss=0.2148, pruned_loss=0.02695, over 1442139.00 frames. ], batch size: 64, lr: 4.83e-03, grad_scale: 8.0 +2023-03-21 07:26:12,934 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 +2023-03-21 07:26:21,169 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 07:26:24,170 INFO [train.py:901] (1/2) Epoch 34, batch 2400, loss[loss=0.1316, simple_loss=0.2148, pruned_loss=0.02417, over 7280.00 frames. ], tot_loss[loss=0.1341, simple_loss=0.2148, pruned_loss=0.02664, over 1441716.34 frames. ], batch size: 57, lr: 4.83e-03, grad_scale: 8.0 +2023-03-21 07:26:27,180 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 07:26:36,113 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95616.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:26:36,700 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95617.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:26:37,586 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 07:26:37,777 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3216, 2.7567, 3.3387, 3.2160, 2.9558, 2.8424, 3.4038, 2.6099], + device='cuda:1'), covar=tensor([0.0360, 0.0388, 0.0593, 0.0596, 0.0616, 0.0769, 0.0531, 0.2026], + device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0340, 0.0276, 0.0360, 0.0295, 0.0291, 0.0350, 0.0258], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:26:40,737 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 07:26:40,801 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95624.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:26:49,096 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.337e+02 1.742e+02 2.021e+02 2.360e+02 4.581e+02, threshold=4.042e+02, percent-clipped=1.0 +2023-03-21 07:26:50,152 INFO [train.py:901] (1/2) Epoch 34, batch 2450, loss[loss=0.1434, simple_loss=0.2309, pruned_loss=0.02798, over 7346.00 frames. ], tot_loss[loss=0.1346, simple_loss=0.2153, pruned_loss=0.02696, over 1441668.44 frames. ], batch size: 73, lr: 4.83e-03, grad_scale: 8.0 +2023-03-21 07:27:06,769 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 07:27:08,422 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95678.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:27:15,809 INFO [train.py:901] (1/2) Epoch 34, batch 2500, loss[loss=0.1264, simple_loss=0.2048, pruned_loss=0.02402, over 7238.00 frames. ], tot_loss[loss=0.1342, simple_loss=0.2151, pruned_loss=0.02669, over 1442420.81 frames. ], batch size: 45, lr: 4.83e-03, grad_scale: 8.0 +2023-03-21 07:27:16,826 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95695.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:27:31,238 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 07:27:31,488 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-03-21 07:27:40,894 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.816e+02 2.143e+02 2.532e+02 3.542e+02, threshold=4.287e+02, percent-clipped=0.0 +2023-03-21 07:27:41,899 INFO [train.py:901] (1/2) Epoch 34, batch 2550, loss[loss=0.1334, simple_loss=0.2103, pruned_loss=0.02826, over 7272.00 frames. ], tot_loss[loss=0.1349, simple_loss=0.2157, pruned_loss=0.02702, over 1441805.92 frames. ], batch size: 77, lr: 4.83e-03, grad_scale: 8.0 +2023-03-21 07:27:51,025 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95761.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 07:27:51,998 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.6460, 5.2268, 5.1015, 5.6136, 5.4441, 5.6079, 5.1528, 5.2384], + device='cuda:1'), covar=tensor([0.0769, 0.1750, 0.1905, 0.0986, 0.0836, 0.1026, 0.0584, 0.1110], + device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0383, 0.0293, 0.0304, 0.0227, 0.0362, 0.0224, 0.0268], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:28:06,902 INFO [train.py:901] (1/2) Epoch 34, batch 2600, loss[loss=0.1108, simple_loss=0.1886, pruned_loss=0.01653, over 7173.00 frames. ], tot_loss[loss=0.1347, simple_loss=0.2156, pruned_loss=0.02693, over 1440655.96 frames. ], batch size: 41, lr: 4.83e-03, grad_scale: 8.0 +2023-03-21 07:28:15,814 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=95809.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 07:28:17,386 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95812.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 07:28:25,034 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-21 07:28:31,405 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.727e+02 2.112e+02 2.461e+02 5.012e+02, threshold=4.224e+02, percent-clipped=2.0 +2023-03-21 07:28:32,387 INFO [train.py:901] (1/2) Epoch 34, batch 2650, loss[loss=0.1398, simple_loss=0.2235, pruned_loss=0.02806, over 7143.00 frames. ], tot_loss[loss=0.1346, simple_loss=0.2155, pruned_loss=0.0269, over 1441968.15 frames. ], batch size: 98, lr: 4.83e-03, grad_scale: 8.0 +2023-03-21 07:28:34,394 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95847.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:28:47,109 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95873.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 07:28:56,648 INFO [train.py:901] (1/2) Epoch 34, batch 2700, loss[loss=0.1322, simple_loss=0.216, pruned_loss=0.02417, over 7275.00 frames. ], tot_loss[loss=0.1344, simple_loss=0.2151, pruned_loss=0.02686, over 1440883.72 frames. ], batch size: 77, lr: 4.83e-03, grad_scale: 8.0 +2023-03-21 07:29:04,725 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95908.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 07:29:08,644 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95916.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:29:12,609 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95924.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:29:20,858 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.273e+02 1.771e+02 2.102e+02 2.649e+02 3.915e+02, threshold=4.205e+02, percent-clipped=0.0 +2023-03-21 07:29:21,982 INFO [train.py:901] (1/2) Epoch 34, batch 2750, loss[loss=0.1435, simple_loss=0.2195, pruned_loss=0.03374, over 7342.00 frames. ], tot_loss[loss=0.1345, simple_loss=0.2152, pruned_loss=0.02691, over 1439891.84 frames. ], batch size: 75, lr: 4.82e-03, grad_scale: 8.0 +2023-03-21 07:29:32,407 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=95964.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:29:36,412 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=95972.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:29:36,903 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95973.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:29:38,867 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0451, 4.5520, 4.6585, 4.4879, 4.5525, 4.0855, 4.5909, 4.5192], + device='cuda:1'), covar=tensor([0.0528, 0.0429, 0.0355, 0.0560, 0.0309, 0.0443, 0.0352, 0.0382], + device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0248, 0.0190, 0.0191, 0.0151, 0.0224, 0.0198, 0.0144], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:29:40,370 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95980.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:29:44,583 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 +2023-03-21 07:29:46,643 INFO [train.py:901] (1/2) Epoch 34, batch 2800, loss[loss=0.1489, simple_loss=0.2323, pruned_loss=0.03274, over 7255.00 frames. ], tot_loss[loss=0.1345, simple_loss=0.215, pruned_loss=0.02698, over 1441640.22 frames. ], batch size: 55, lr: 4.82e-03, grad_scale: 8.0 +2023-03-21 07:29:47,708 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95995.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:29:49,188 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8049, 2.8270, 3.3669, 3.0137, 3.1740, 3.0976, 2.6024, 3.2957], + device='cuda:1'), covar=tensor([0.2334, 0.0843, 0.0937, 0.1413, 0.1005, 0.0943, 0.1859, 0.1029], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0065, 0.0049, 0.0048, 0.0047, 0.0048, 0.0066, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 07:30:15,784 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 07:30:15,839 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 07:30:15,862 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 07:30:15,879 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 07:30:16,337 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 07:30:16,352 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 07:30:16,362 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 07:30:16,458 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 07:30:16,536 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 07:30:16,551 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 07:30:16,567 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 07:30:16,727 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 07:30:16,732 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 07:30:16,740 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 07:30:16,878 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 07:30:16,882 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 07:30:17,254 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 07:30:24,327 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 +2023-03-21 07:30:24,383 INFO [train.py:901] (1/2) Epoch 35, batch 0, loss[loss=0.1146, simple_loss=0.1962, pruned_loss=0.01654, over 7305.00 frames. ], tot_loss[loss=0.1146, simple_loss=0.1962, pruned_loss=0.01654, over 7305.00 frames. ], batch size: 68, lr: 4.75e-03, grad_scale: 8.0 +2023-03-21 07:30:24,384 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 07:30:50,187 INFO [train.py:935] (1/2) Epoch 35, validation: loss=0.1643, simple_loss=0.2559, pruned_loss=0.03642, over 1622729.00 frames. +2023-03-21 07:30:50,187 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 07:30:56,261 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 07:31:02,946 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.206e+02 1.782e+02 1.999e+02 2.365e+02 3.842e+02, threshold=3.998e+02, percent-clipped=0.0 +2023-03-21 07:31:03,130 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96041.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:31:04,031 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=96043.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:31:07,475 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 07:31:15,065 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 07:31:16,030 INFO [train.py:901] (1/2) Epoch 35, batch 50, loss[loss=0.1519, simple_loss=0.2328, pruned_loss=0.0355, over 7362.00 frames. ], tot_loss[loss=0.1317, simple_loss=0.2134, pruned_loss=0.02499, over 325436.33 frames. ], batch size: 63, lr: 4.75e-03, grad_scale: 8.0 +2023-03-21 07:31:17,102 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 07:31:20,610 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 07:31:22,828 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96080.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:31:29,551 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2823, 2.8219, 2.1416, 3.0257, 3.2019, 3.2105, 2.7730, 2.7915], + device='cuda:1'), covar=tensor([0.2226, 0.1022, 0.3483, 0.0917, 0.0310, 0.0253, 0.0384, 0.0345], + device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0233, 0.0249, 0.0261, 0.0194, 0.0194, 0.0213, 0.0223], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 07:31:37,955 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 07:31:38,448 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 07:31:41,925 INFO [train.py:901] (1/2) Epoch 35, batch 100, loss[loss=0.1233, simple_loss=0.2007, pruned_loss=0.02292, over 7205.00 frames. ], tot_loss[loss=0.131, simple_loss=0.2119, pruned_loss=0.02506, over 574100.31 frames. ], batch size: 39, lr: 4.75e-03, grad_scale: 8.0 +2023-03-21 07:31:54,359 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.190e+02 1.736e+02 2.019e+02 2.220e+02 3.801e+02, threshold=4.038e+02, percent-clipped=0.0 +2023-03-21 07:31:54,580 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96141.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:31:56,245 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 +2023-03-21 07:32:06,603 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96165.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:32:07,444 INFO [train.py:901] (1/2) Epoch 35, batch 150, loss[loss=0.1529, simple_loss=0.2397, pruned_loss=0.03303, over 7138.00 frames. ], tot_loss[loss=0.1322, simple_loss=0.2127, pruned_loss=0.0258, over 765483.99 frames. ], batch size: 98, lr: 4.75e-03, grad_scale: 8.0 +2023-03-21 07:32:08,035 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96168.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 07:32:26,416 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96203.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 07:32:33,899 INFO [train.py:901] (1/2) Epoch 35, batch 200, loss[loss=0.1421, simple_loss=0.2189, pruned_loss=0.03266, over 7273.00 frames. ], tot_loss[loss=0.132, simple_loss=0.2122, pruned_loss=0.0259, over 914539.11 frames. ], batch size: 52, lr: 4.75e-03, grad_scale: 8.0 +2023-03-21 07:32:38,637 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96226.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:32:41,064 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 07:32:45,115 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 07:32:46,087 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.234e+02 1.725e+02 2.081e+02 2.460e+02 5.560e+02, threshold=4.161e+02, percent-clipped=1.0 +2023-03-21 07:32:52,138 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 07:32:59,857 INFO [train.py:901] (1/2) Epoch 35, batch 250, loss[loss=0.1314, simple_loss=0.2133, pruned_loss=0.02479, over 7270.00 frames. ], tot_loss[loss=0.1324, simple_loss=0.2126, pruned_loss=0.02604, over 1031360.85 frames. ], batch size: 47, lr: 4.75e-03, grad_scale: 8.0 +2023-03-21 07:33:03,162 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96273.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:33:05,567 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 07:33:25,521 INFO [train.py:901] (1/2) Epoch 35, batch 300, loss[loss=0.1236, simple_loss=0.2043, pruned_loss=0.02141, over 7228.00 frames. ], tot_loss[loss=0.133, simple_loss=0.2136, pruned_loss=0.02623, over 1123252.34 frames. ], batch size: 45, lr: 4.75e-03, grad_scale: 8.0 +2023-03-21 07:33:26,006 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 07:33:27,585 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=96321.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:33:29,218 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96324.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:33:35,155 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96336.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:33:35,581 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 07:33:37,565 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.756e+02 2.059e+02 2.406e+02 3.966e+02, threshold=4.119e+02, percent-clipped=0.0 +2023-03-21 07:33:50,933 INFO [train.py:901] (1/2) Epoch 35, batch 350, loss[loss=0.1256, simple_loss=0.2129, pruned_loss=0.01917, over 7326.00 frames. ], tot_loss[loss=0.1324, simple_loss=0.213, pruned_loss=0.02589, over 1194376.43 frames. ], batch size: 54, lr: 4.74e-03, grad_scale: 8.0 +2023-03-21 07:33:59,810 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96385.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:34:02,288 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96390.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:34:09,989 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 07:34:16,493 INFO [train.py:901] (1/2) Epoch 35, batch 400, loss[loss=0.1385, simple_loss=0.2229, pruned_loss=0.02707, over 7315.00 frames. ], tot_loss[loss=0.1329, simple_loss=0.2137, pruned_loss=0.02606, over 1249503.85 frames. ], batch size: 49, lr: 4.74e-03, grad_scale: 8.0 +2023-03-21 07:34:26,177 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96436.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:34:28,583 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.391e+02 1.843e+02 2.067e+02 2.365e+02 4.098e+02, threshold=4.134e+02, percent-clipped=0.0 +2023-03-21 07:34:34,245 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96451.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:34:42,144 INFO [train.py:901] (1/2) Epoch 35, batch 450, loss[loss=0.1371, simple_loss=0.2122, pruned_loss=0.03102, over 7333.00 frames. ], tot_loss[loss=0.1331, simple_loss=0.2135, pruned_loss=0.02633, over 1289115.54 frames. ], batch size: 49, lr: 4.74e-03, grad_scale: 8.0 +2023-03-21 07:34:42,768 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96468.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 07:34:51,695 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 07:34:52,212 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 07:35:00,857 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96503.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:35:07,383 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=96516.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 07:35:07,797 INFO [train.py:901] (1/2) Epoch 35, batch 500, loss[loss=0.1355, simple_loss=0.2196, pruned_loss=0.02568, over 7261.00 frames. ], tot_loss[loss=0.1334, simple_loss=0.2139, pruned_loss=0.0265, over 1322015.05 frames. ], batch size: 89, lr: 4.74e-03, grad_scale: 8.0 +2023-03-21 07:35:09,824 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96521.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:35:20,462 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.185e+02 1.758e+02 2.053e+02 2.649e+02 4.888e+02, threshold=4.106e+02, percent-clipped=1.0 +2023-03-21 07:35:23,978 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3435, 4.1473, 3.6001, 3.8034, 3.5134, 2.3987, 1.9867, 4.3113], + device='cuda:1'), covar=tensor([0.0050, 0.0067, 0.0124, 0.0068, 0.0115, 0.0531, 0.0596, 0.0044], + device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0090, 0.0111, 0.0095, 0.0126, 0.0133, 0.0130, 0.0103], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 07:35:24,175 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 +2023-03-21 07:35:24,397 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 07:35:25,424 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=96551.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:35:25,880 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 07:35:26,851 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 07:35:28,354 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 07:35:33,901 INFO [train.py:901] (1/2) Epoch 35, batch 550, loss[loss=0.1231, simple_loss=0.2048, pruned_loss=0.0207, over 7262.00 frames. ], tot_loss[loss=0.1339, simple_loss=0.2146, pruned_loss=0.02661, over 1350428.95 frames. ], batch size: 55, lr: 4.74e-03, grad_scale: 8.0 +2023-03-21 07:35:33,915 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 07:35:43,580 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 +2023-03-21 07:35:43,733 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-03-21 07:35:44,377 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 07:35:44,959 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2523, 3.8238, 3.9566, 3.9270, 3.9024, 3.8243, 4.1308, 3.6339], + device='cuda:1'), covar=tensor([0.0118, 0.0184, 0.0117, 0.0183, 0.0442, 0.0113, 0.0141, 0.0177], + device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0101, 0.0100, 0.0089, 0.0174, 0.0107, 0.0105, 0.0110], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:35:47,947 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.1743, 4.6896, 4.5238, 5.0970, 4.9518, 5.0718, 4.4766, 4.6413], + device='cuda:1'), covar=tensor([0.0821, 0.2385, 0.2310, 0.1120, 0.0982, 0.1109, 0.0798, 0.1064], + device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0379, 0.0291, 0.0300, 0.0224, 0.0358, 0.0221, 0.0266], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:35:52,735 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 07:35:55,687 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 07:35:59,820 INFO [train.py:901] (1/2) Epoch 35, batch 600, loss[loss=0.1362, simple_loss=0.2216, pruned_loss=0.02538, over 7299.00 frames. ], tot_loss[loss=0.1335, simple_loss=0.2145, pruned_loss=0.0263, over 1370758.86 frames. ], batch size: 57, lr: 4.74e-03, grad_scale: 8.0 +2023-03-21 07:36:03,282 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 07:36:09,431 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96636.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:36:11,805 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.247e+02 1.802e+02 2.034e+02 2.397e+02 4.474e+02, threshold=4.067e+02, percent-clipped=1.0 +2023-03-21 07:36:19,390 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 07:36:25,457 INFO [train.py:901] (1/2) Epoch 35, batch 650, loss[loss=0.1252, simple_loss=0.201, pruned_loss=0.02467, over 7335.00 frames. ], tot_loss[loss=0.1335, simple_loss=0.2145, pruned_loss=0.02621, over 1386914.05 frames. ], batch size: 75, lr: 4.74e-03, grad_scale: 8.0 +2023-03-21 07:36:27,953 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 07:36:32,102 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96680.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:36:34,077 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=96684.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:36:46,085 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 07:36:51,024 INFO [train.py:901] (1/2) Epoch 35, batch 700, loss[loss=0.135, simple_loss=0.2182, pruned_loss=0.02596, over 7268.00 frames. ], tot_loss[loss=0.133, simple_loss=0.2141, pruned_loss=0.026, over 1399319.46 frames. ], batch size: 52, lr: 4.74e-03, grad_scale: 8.0 +2023-03-21 07:36:54,519 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 07:37:01,199 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96736.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:37:03,569 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.274e+02 1.655e+02 1.913e+02 2.352e+02 3.396e+02, threshold=3.826e+02, percent-clipped=0.0 +2023-03-21 07:37:06,175 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96746.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:37:16,484 INFO [train.py:901] (1/2) Epoch 35, batch 750, loss[loss=0.1422, simple_loss=0.2287, pruned_loss=0.0279, over 7252.00 frames. ], tot_loss[loss=0.1333, simple_loss=0.2142, pruned_loss=0.02618, over 1410578.70 frames. ], batch size: 55, lr: 4.73e-03, grad_scale: 8.0 +2023-03-21 07:37:19,013 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 07:37:19,511 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 07:37:21,186 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96776.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:37:25,064 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=96784.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:37:33,558 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 07:37:38,796 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 07:37:39,694 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-21 07:37:42,330 INFO [train.py:901] (1/2) Epoch 35, batch 800, loss[loss=0.1292, simple_loss=0.2101, pruned_loss=0.02413, over 7311.00 frames. ], tot_loss[loss=0.134, simple_loss=0.215, pruned_loss=0.02646, over 1417719.36 frames. ], batch size: 80, lr: 4.73e-03, grad_scale: 8.0 +2023-03-21 07:37:45,029 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96821.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:37:45,396 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 07:37:46,409 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 07:37:52,968 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96837.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:37:54,835 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.675e+02 2.012e+02 2.364e+02 4.591e+02, threshold=4.024e+02, percent-clipped=3.0 +2023-03-21 07:37:57,396 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 07:38:07,909 INFO [train.py:901] (1/2) Epoch 35, batch 850, loss[loss=0.1269, simple_loss=0.2123, pruned_loss=0.02069, over 7307.00 frames. ], tot_loss[loss=0.1347, simple_loss=0.2156, pruned_loss=0.02688, over 1423136.90 frames. ], batch size: 49, lr: 4.73e-03, grad_scale: 8.0 +2023-03-21 07:38:08,950 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=96869.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:38:14,966 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 07:38:15,435 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 07:38:17,166 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7269, 2.3561, 2.9539, 2.8050, 2.7991, 2.7997, 2.3485, 2.8839], + device='cuda:1'), covar=tensor([0.1676, 0.1163, 0.1379, 0.1076, 0.1151, 0.0968, 0.2557, 0.1529], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0065, 0.0049, 0.0048, 0.0047, 0.0049, 0.0065, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 07:38:21,671 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 07:38:24,643 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 07:38:34,189 INFO [train.py:901] (1/2) Epoch 35, batch 900, loss[loss=0.1217, simple_loss=0.1976, pruned_loss=0.02289, over 7149.00 frames. ], tot_loss[loss=0.1344, simple_loss=0.2154, pruned_loss=0.02666, over 1427413.69 frames. ], batch size: 41, lr: 4.73e-03, grad_scale: 8.0 +2023-03-21 07:38:44,844 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0322, 4.5826, 4.4045, 5.0114, 4.8390, 4.9666, 4.4931, 4.5959], + device='cuda:1'), covar=tensor([0.0762, 0.2177, 0.2165, 0.0861, 0.0914, 0.1113, 0.0651, 0.1007], + device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0378, 0.0292, 0.0300, 0.0222, 0.0357, 0.0220, 0.0264], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:38:46,262 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.358e+02 1.734e+02 2.102e+02 2.438e+02 4.011e+02, threshold=4.204e+02, percent-clipped=0.0 +2023-03-21 07:38:59,393 INFO [train.py:901] (1/2) Epoch 35, batch 950, loss[loss=0.1366, simple_loss=0.2228, pruned_loss=0.0252, over 7294.00 frames. ], tot_loss[loss=0.1341, simple_loss=0.2154, pruned_loss=0.02639, over 1432093.44 frames. ], batch size: 70, lr: 4.73e-03, grad_scale: 8.0 +2023-03-21 07:39:01,517 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 07:39:06,598 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96980.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:39:11,540 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7491, 2.7439, 3.8182, 3.5125, 3.6341, 3.6513, 3.7858, 3.3696], + device='cuda:1'), covar=tensor([0.0044, 0.0213, 0.0045, 0.0063, 0.0051, 0.0057, 0.0064, 0.0088], + device='cuda:1'), in_proj_covar=tensor([0.0046, 0.0066, 0.0055, 0.0053, 0.0053, 0.0058, 0.0047, 0.0074], + device='cuda:1'), out_proj_covar=tensor([7.9165e-05, 1.3715e-04, 1.0258e-04, 9.3074e-05, 9.2426e-05, 1.0317e-04, + 9.1643e-05, 1.3903e-04], device='cuda:1') +2023-03-21 07:39:17,609 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97000.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:39:24,959 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1541, 2.2990, 2.6589, 2.2473, 2.4127, 2.5565, 2.1585, 2.2581], + device='cuda:1'), covar=tensor([0.0549, 0.0516, 0.0215, 0.0331, 0.0673, 0.0676, 0.0361, 0.0267], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0035, 0.0037, 0.0036, 0.0034, 0.0034, 0.0039, 0.0039], + device='cuda:1'), out_proj_covar=tensor([9.3924e-05, 9.2528e-05, 9.3578e-05, 8.9979e-05, 9.0013e-05, 8.7980e-05, + 9.6644e-05, 9.8082e-05], device='cuda:1') +2023-03-21 07:39:25,836 INFO [train.py:901] (1/2) Epoch 35, batch 1000, loss[loss=0.1363, simple_loss=0.223, pruned_loss=0.0248, over 7259.00 frames. ], tot_loss[loss=0.1338, simple_loss=0.215, pruned_loss=0.02635, over 1434718.59 frames. ], batch size: 89, lr: 4.73e-03, grad_scale: 16.0 +2023-03-21 07:39:25,846 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 07:39:31,476 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=97028.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:39:38,018 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.712e+02 1.972e+02 2.279e+02 4.692e+02, threshold=3.943e+02, percent-clipped=1.0 +2023-03-21 07:39:40,669 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97046.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:39:46,229 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 07:39:48,909 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97061.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:39:51,809 INFO [train.py:901] (1/2) Epoch 35, batch 1050, loss[loss=0.1098, simple_loss=0.183, pruned_loss=0.01823, over 7034.00 frames. ], tot_loss[loss=0.1338, simple_loss=0.215, pruned_loss=0.02629, over 1434755.94 frames. ], batch size: 35, lr: 4.73e-03, grad_scale: 16.0 +2023-03-21 07:40:05,941 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=97094.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:40:07,868 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 07:40:11,875 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 07:40:15,042 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4606, 1.7268, 1.4749, 1.6336, 1.7588, 1.4989, 1.6312, 1.2744], + device='cuda:1'), covar=tensor([0.0159, 0.0138, 0.0251, 0.0190, 0.0099, 0.0148, 0.0256, 0.0198], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0033, 0.0033, 0.0035, 0.0033, 0.0032, 0.0035, 0.0041], + device='cuda:1'), out_proj_covar=tensor([3.9976e-05, 3.6738e-05, 3.7150e-05, 3.8405e-05, 3.6921e-05, 3.5206e-05, + 3.9752e-05, 4.5551e-05], device='cuda:1') +2023-03-21 07:40:17,344 INFO [train.py:901] (1/2) Epoch 35, batch 1100, loss[loss=0.1313, simple_loss=0.2081, pruned_loss=0.0272, over 7294.00 frames. ], tot_loss[loss=0.1337, simple_loss=0.2151, pruned_loss=0.02617, over 1438517.30 frames. ], batch size: 66, lr: 4.73e-03, grad_scale: 16.0 +2023-03-21 07:40:24,935 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97132.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:40:29,388 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.690e+02 1.951e+02 2.274e+02 4.378e+02, threshold=3.903e+02, percent-clipped=1.0 +2023-03-21 07:40:35,131 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0576, 3.8996, 3.3575, 3.6170, 3.1516, 2.3660, 1.9074, 4.1056], + device='cuda:1'), covar=tensor([0.0052, 0.0066, 0.0131, 0.0072, 0.0141, 0.0555, 0.0616, 0.0048], + device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0089, 0.0110, 0.0093, 0.0124, 0.0132, 0.0128, 0.0102], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 07:40:40,475 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 07:40:40,951 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 07:40:42,907 INFO [train.py:901] (1/2) Epoch 35, batch 1150, loss[loss=0.1437, simple_loss=0.2205, pruned_loss=0.0335, over 7288.00 frames. ], tot_loss[loss=0.1336, simple_loss=0.2148, pruned_loss=0.02619, over 1440881.89 frames. ], batch size: 66, lr: 4.73e-03, grad_scale: 16.0 +2023-03-21 07:40:54,038 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 07:40:54,539 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 07:40:57,244 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3294, 2.1993, 2.4570, 2.9022, 2.5508, 2.6308, 2.5475, 2.6151], + device='cuda:1'), covar=tensor([0.1626, 0.3538, 0.2568, 0.1206, 0.3083, 0.2001, 0.2039, 0.4802], + device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0077, 0.0067, 0.0061, 0.0059, 0.0061, 0.0103, 0.0063], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:41:08,765 INFO [train.py:901] (1/2) Epoch 35, batch 1200, loss[loss=0.1387, simple_loss=0.2201, pruned_loss=0.02867, over 7324.00 frames. ], tot_loss[loss=0.1334, simple_loss=0.2145, pruned_loss=0.0261, over 1441742.83 frames. ], batch size: 59, lr: 4.72e-03, grad_scale: 16.0 +2023-03-21 07:41:21,318 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 +2023-03-21 07:41:21,384 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.240e+02 1.833e+02 2.092e+02 2.573e+02 3.839e+02, threshold=4.184e+02, percent-clipped=0.0 +2023-03-21 07:41:23,498 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6486, 3.0053, 2.6414, 2.9140, 2.7947, 2.4687, 2.9329, 2.6244], + device='cuda:1'), covar=tensor([0.0828, 0.0680, 0.1087, 0.0718, 0.0932, 0.0824, 0.0799, 0.1305], + device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0059, 0.0068, 0.0059, 0.0056, 0.0060, 0.0057, 0.0054], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:41:26,305 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 07:41:31,435 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97260.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:41:34,828 INFO [train.py:901] (1/2) Epoch 35, batch 1250, loss[loss=0.1179, simple_loss=0.2036, pruned_loss=0.01611, over 7167.00 frames. ], tot_loss[loss=0.1334, simple_loss=0.2145, pruned_loss=0.02616, over 1441119.84 frames. ], batch size: 41, lr: 4.72e-03, grad_scale: 16.0 +2023-03-21 07:41:35,977 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8841, 4.0464, 3.8543, 4.0679, 3.6483, 3.9332, 4.2443, 4.2761], + device='cuda:1'), covar=tensor([0.0203, 0.0142, 0.0201, 0.0152, 0.0378, 0.0302, 0.0244, 0.0188], + device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0120, 0.0113, 0.0115, 0.0107, 0.0098, 0.0094, 0.0094], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:41:40,283 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-21 07:41:49,380 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 07:41:53,971 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 07:41:55,523 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 07:42:00,727 INFO [train.py:901] (1/2) Epoch 35, batch 1300, loss[loss=0.1517, simple_loss=0.2398, pruned_loss=0.03184, over 7145.00 frames. ], tot_loss[loss=0.1339, simple_loss=0.215, pruned_loss=0.02641, over 1440562.43 frames. ], batch size: 98, lr: 4.72e-03, grad_scale: 16.0 +2023-03-21 07:42:02,971 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97321.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:42:03,626 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-03-21 07:42:10,068 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3027, 1.6201, 1.4397, 1.4435, 1.5082, 1.4743, 1.4539, 1.2827], + device='cuda:1'), covar=tensor([0.0208, 0.0124, 0.0179, 0.0136, 0.0142, 0.0131, 0.0121, 0.0139], + device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0034, 0.0033, 0.0035, 0.0034, 0.0032, 0.0035, 0.0042], + device='cuda:1'), out_proj_covar=tensor([4.0422e-05, 3.7216e-05, 3.7602e-05, 3.8546e-05, 3.7680e-05, 3.5783e-05, + 3.9998e-05, 4.6492e-05], device='cuda:1') +2023-03-21 07:42:13,453 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 1.729e+02 1.997e+02 2.434e+02 4.053e+02, threshold=3.995e+02, percent-clipped=0.0 +2023-03-21 07:42:14,066 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97342.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:42:16,597 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97347.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:42:18,995 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 07:42:21,013 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 07:42:21,082 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97356.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:42:25,016 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 07:42:26,294 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-03-21 07:42:26,537 INFO [train.py:901] (1/2) Epoch 35, batch 1350, loss[loss=0.129, simple_loss=0.2104, pruned_loss=0.02385, over 7280.00 frames. ], tot_loss[loss=0.1342, simple_loss=0.2155, pruned_loss=0.02648, over 1441665.65 frames. ], batch size: 52, lr: 4.72e-03, grad_scale: 16.0 +2023-03-21 07:42:30,234 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97374.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:42:35,123 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 07:42:45,973 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97403.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 07:42:48,430 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97408.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:42:52,844 INFO [train.py:901] (1/2) Epoch 35, batch 1400, loss[loss=0.1438, simple_loss=0.2251, pruned_loss=0.0312, over 7355.00 frames. ], tot_loss[loss=0.1336, simple_loss=0.2147, pruned_loss=0.02629, over 1439161.15 frames. ], batch size: 63, lr: 4.72e-03, grad_scale: 16.0 +2023-03-21 07:43:01,109 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97432.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:43:02,682 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97435.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:43:03,733 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9803, 3.1892, 2.4484, 3.8145, 2.6378, 3.2312, 1.7153, 2.4121], + device='cuda:1'), covar=tensor([0.0402, 0.0685, 0.2189, 0.0548, 0.0426, 0.0464, 0.3612, 0.1764], + device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0261, 0.0286, 0.0272, 0.0274, 0.0268, 0.0240, 0.0262], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 07:43:05,157 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6529, 3.7976, 3.6516, 3.7764, 3.3703, 3.6971, 4.0332, 4.0205], + device='cuda:1'), covar=tensor([0.0237, 0.0163, 0.0206, 0.0189, 0.0388, 0.0344, 0.0233, 0.0213], + device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0122, 0.0116, 0.0119, 0.0110, 0.0101, 0.0096, 0.0097], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:43:06,069 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.642e+02 1.848e+02 2.311e+02 4.158e+02, threshold=3.695e+02, percent-clipped=1.0 +2023-03-21 07:43:10,601 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 07:43:18,611 INFO [train.py:901] (1/2) Epoch 35, batch 1450, loss[loss=0.1291, simple_loss=0.2193, pruned_loss=0.01945, over 7354.00 frames. ], tot_loss[loss=0.1338, simple_loss=0.2148, pruned_loss=0.02646, over 1441013.24 frames. ], batch size: 73, lr: 4.72e-03, grad_scale: 8.0 +2023-03-21 07:43:25,233 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=97480.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:43:26,781 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6578, 3.6739, 3.5718, 3.6915, 3.5699, 3.3087, 3.7110, 3.7906], + device='cuda:1'), covar=tensor([0.0390, 0.0315, 0.0364, 0.0456, 0.0428, 0.0547, 0.0492, 0.0425], + device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0123, 0.0116, 0.0119, 0.0110, 0.0101, 0.0096, 0.0097], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:43:34,277 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 07:43:35,833 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0455, 4.6020, 4.3459, 4.9504, 4.7868, 4.9532, 4.3698, 4.5592], + device='cuda:1'), covar=tensor([0.0785, 0.2132, 0.2274, 0.1079, 0.0970, 0.1015, 0.0681, 0.1181], + device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0377, 0.0291, 0.0297, 0.0221, 0.0354, 0.0217, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:43:40,403 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7448, 2.9172, 2.2282, 3.4929, 2.1221, 2.9525, 1.5410, 2.2975], + device='cuda:1'), covar=tensor([0.0440, 0.0890, 0.2863, 0.0669, 0.0386, 0.0707, 0.4000, 0.1760], + device='cuda:1'), in_proj_covar=tensor([0.0285, 0.0261, 0.0285, 0.0272, 0.0274, 0.0268, 0.0240, 0.0263], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 07:43:44,138 INFO [train.py:901] (1/2) Epoch 35, batch 1500, loss[loss=0.1246, simple_loss=0.2063, pruned_loss=0.02147, over 7303.00 frames. ], tot_loss[loss=0.1341, simple_loss=0.2147, pruned_loss=0.02674, over 1442234.67 frames. ], batch size: 80, lr: 4.72e-03, grad_scale: 8.0 +2023-03-21 07:43:50,302 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 07:43:57,133 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.381e+02 1.775e+02 2.183e+02 2.591e+02 6.177e+02, threshold=4.365e+02, percent-clipped=2.0 +2023-03-21 07:44:03,844 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97555.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:44:04,827 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9481, 4.4738, 4.4685, 4.4676, 4.4687, 4.0032, 4.4991, 4.4064], + device='cuda:1'), covar=tensor([0.0528, 0.0429, 0.0473, 0.0531, 0.0343, 0.0437, 0.0436, 0.0431], + device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0247, 0.0192, 0.0192, 0.0152, 0.0225, 0.0200, 0.0146], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:44:09,681 INFO [train.py:901] (1/2) Epoch 35, batch 1550, loss[loss=0.1535, simple_loss=0.2357, pruned_loss=0.03569, over 6686.00 frames. ], tot_loss[loss=0.1337, simple_loss=0.2145, pruned_loss=0.02648, over 1441817.72 frames. ], batch size: 107, lr: 4.72e-03, grad_scale: 8.0 +2023-03-21 07:44:13,290 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 07:44:13,561 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-03-21 07:44:23,445 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8533, 3.2505, 3.7674, 3.8053, 3.8538, 3.9600, 3.9512, 3.7588], + device='cuda:1'), covar=tensor([0.0029, 0.0108, 0.0031, 0.0031, 0.0029, 0.0026, 0.0037, 0.0051], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0065, 0.0054, 0.0052, 0.0052, 0.0057, 0.0045, 0.0072], + device='cuda:1'), out_proj_covar=tensor([7.8180e-05, 1.3426e-04, 1.0061e-04, 9.2120e-05, 9.0452e-05, 1.0077e-04, + 8.8169e-05, 1.3600e-04], device='cuda:1') +2023-03-21 07:44:35,932 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:44:36,020 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:44:36,385 INFO [train.py:901] (1/2) Epoch 35, batch 1600, loss[loss=0.1172, simple_loss=0.1962, pruned_loss=0.01915, over 7349.00 frames. ], tot_loss[loss=0.1332, simple_loss=0.2142, pruned_loss=0.02612, over 1440029.34 frames. ], batch size: 44, lr: 4.71e-03, grad_scale: 8.0 +2023-03-21 07:44:44,853 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 07:44:45,459 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2262, 2.0939, 2.5195, 2.1787, 2.3681, 2.3276, 2.2043, 1.7704], + device='cuda:1'), covar=tensor([0.0387, 0.0475, 0.0220, 0.0320, 0.0388, 0.0374, 0.0284, 0.0367], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0037, 0.0037, 0.0036, 0.0034, 0.0035, 0.0040, 0.0039], + device='cuda:1'), out_proj_covar=tensor([9.6473e-05, 9.4996e-05, 9.4197e-05, 9.1548e-05, 9.1118e-05, 9.0174e-05, + 9.8189e-05, 9.9561e-05], device='cuda:1') +2023-03-21 07:44:45,834 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 07:44:48,738 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.345e+02 1.881e+02 2.141e+02 2.690e+02 4.588e+02, threshold=4.282e+02, percent-clipped=0.0 +2023-03-21 07:44:48,775 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 07:44:55,902 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97656.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:44:57,833 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 07:45:01,965 INFO [train.py:901] (1/2) Epoch 35, batch 1650, loss[loss=0.1348, simple_loss=0.2154, pruned_loss=0.02711, over 7322.00 frames. ], tot_loss[loss=0.1335, simple_loss=0.2144, pruned_loss=0.02628, over 1440738.79 frames. ], batch size: 59, lr: 4.71e-03, grad_scale: 8.0 +2023-03-21 07:45:02,473 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 07:45:05,644 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97674.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 07:45:10,540 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 07:45:18,328 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97698.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 07:45:20,774 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97703.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:45:21,262 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=97704.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:45:27,134 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 07:45:27,639 INFO [train.py:901] (1/2) Epoch 35, batch 1700, loss[loss=0.1248, simple_loss=0.2138, pruned_loss=0.01787, over 7285.00 frames. ], tot_loss[loss=0.1337, simple_loss=0.2146, pruned_loss=0.02637, over 1441489.98 frames. ], batch size: 66, lr: 4.71e-03, grad_scale: 8.0 +2023-03-21 07:45:31,167 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 07:45:34,236 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97730.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:45:36,778 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97735.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 07:45:36,788 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4075, 3.4823, 2.0737, 3.6856, 3.8913, 3.9676, 3.4188, 3.3681], + device='cuda:1'), covar=tensor([0.2293, 0.0682, 0.4567, 0.0364, 0.0197, 0.0203, 0.0396, 0.0418], + device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0232, 0.0248, 0.0256, 0.0193, 0.0193, 0.0212, 0.0220], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 07:45:40,191 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 1.701e+02 2.015e+02 2.345e+02 4.849e+02, threshold=4.029e+02, percent-clipped=2.0 +2023-03-21 07:45:41,246 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 07:45:45,547 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6667, 2.0011, 1.6451, 1.8101, 1.8795, 1.6298, 1.7630, 1.6633], + device='cuda:1'), covar=tensor([0.0161, 0.0162, 0.0271, 0.0171, 0.0132, 0.0248, 0.0176, 0.0199], + device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0033, 0.0032, 0.0034, 0.0033, 0.0032, 0.0035, 0.0042], + device='cuda:1'), out_proj_covar=tensor([3.9765e-05, 3.6692e-05, 3.6949e-05, 3.7886e-05, 3.6780e-05, 3.5235e-05, + 3.9566e-05, 4.6176e-05], device='cuda:1') +2023-03-21 07:45:53,429 INFO [train.py:901] (1/2) Epoch 35, batch 1750, loss[loss=0.1418, simple_loss=0.2252, pruned_loss=0.02918, over 6831.00 frames. ], tot_loss[loss=0.133, simple_loss=0.214, pruned_loss=0.02601, over 1441129.86 frames. ], batch size: 107, lr: 4.71e-03, grad_scale: 8.0 +2023-03-21 07:45:53,576 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7674, 1.5362, 2.0121, 2.3008, 2.1012, 2.0121, 1.8608, 2.0594], + device='cuda:1'), covar=tensor([0.3213, 0.4808, 0.1300, 0.1169, 0.1970, 0.1607, 0.2655, 0.2515], + device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0075, 0.0066, 0.0060, 0.0058, 0.0060, 0.0100, 0.0061], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:45:58,471 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-21 07:46:06,153 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 07:46:07,161 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 07:46:19,582 INFO [train.py:901] (1/2) Epoch 35, batch 1800, loss[loss=0.1471, simple_loss=0.2258, pruned_loss=0.03419, over 7298.00 frames. ], tot_loss[loss=0.1328, simple_loss=0.2139, pruned_loss=0.02588, over 1443738.71 frames. ], batch size: 80, lr: 4.71e-03, grad_scale: 8.0 +2023-03-21 07:46:29,750 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 07:46:30,529 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-03-21 07:46:32,760 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.185e+02 1.651e+02 2.008e+02 2.380e+02 5.127e+02, threshold=4.015e+02, percent-clipped=1.0 +2023-03-21 07:46:44,440 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 07:46:45,915 INFO [train.py:901] (1/2) Epoch 35, batch 1850, loss[loss=0.1038, simple_loss=0.1739, pruned_loss=0.01688, over 7020.00 frames. ], tot_loss[loss=0.1323, simple_loss=0.2134, pruned_loss=0.02561, over 1444104.91 frames. ], batch size: 35, lr: 4.71e-03, grad_scale: 8.0 +2023-03-21 07:46:54,448 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 07:46:56,427 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97888.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:47:07,854 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97911.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:47:10,354 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97916.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:47:10,769 INFO [train.py:901] (1/2) Epoch 35, batch 1900, loss[loss=0.1399, simple_loss=0.2228, pruned_loss=0.02844, over 7348.00 frames. ], tot_loss[loss=0.1319, simple_loss=0.2129, pruned_loss=0.02552, over 1445248.25 frames. ], batch size: 73, lr: 4.71e-03, grad_scale: 8.0 +2023-03-21 07:47:11,312 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 07:47:24,070 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.234e+02 1.816e+02 2.128e+02 2.562e+02 3.920e+02, threshold=4.257e+02, percent-clipped=0.0 +2023-03-21 07:47:28,349 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97949.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:47:35,764 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=97964.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:47:36,217 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 07:47:37,197 INFO [train.py:901] (1/2) Epoch 35, batch 1950, loss[loss=0.1298, simple_loss=0.2113, pruned_loss=0.02418, over 7324.00 frames. ], tot_loss[loss=0.1328, simple_loss=0.2136, pruned_loss=0.02597, over 1445093.43 frames. ], batch size: 75, lr: 4.71e-03, grad_scale: 8.0 +2023-03-21 07:47:37,553 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-03-21 07:47:47,477 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 07:47:51,566 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 07:47:52,549 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 07:47:53,150 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97998.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:47:56,034 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98003.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:48:01,184 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1550, 3.2375, 2.3300, 3.7598, 2.9112, 3.3153, 1.7117, 2.4658], + device='cuda:1'), covar=tensor([0.0444, 0.0791, 0.2651, 0.0723, 0.0455, 0.0626, 0.3574, 0.1678], + device='cuda:1'), in_proj_covar=tensor([0.0282, 0.0260, 0.0283, 0.0270, 0.0273, 0.0265, 0.0238, 0.0261], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 07:48:03,657 INFO [train.py:901] (1/2) Epoch 35, batch 2000, loss[loss=0.1458, simple_loss=0.2271, pruned_loss=0.03228, over 6686.00 frames. ], tot_loss[loss=0.1328, simple_loss=0.2137, pruned_loss=0.02593, over 1444636.26 frames. ], batch size: 106, lr: 4.70e-03, grad_scale: 8.0 +2023-03-21 07:48:09,309 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2420, 3.5243, 3.1232, 3.4688, 3.3513, 2.9594, 3.4219, 3.0146], + device='cuda:1'), covar=tensor([0.0845, 0.0543, 0.0697, 0.0454, 0.1295, 0.0553, 0.0875, 0.1154], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0057, 0.0066, 0.0057, 0.0054, 0.0059, 0.0056, 0.0054], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:48:10,348 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 07:48:10,915 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98030.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 07:48:10,935 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98030.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:48:16,923 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.660e+02 1.963e+02 2.436e+02 6.666e+02, threshold=3.925e+02, percent-clipped=3.0 +2023-03-21 07:48:19,029 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=98046.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:48:20,952 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 07:48:21,496 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=98051.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:48:28,915 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 07:48:29,417 INFO [train.py:901] (1/2) Epoch 35, batch 2050, loss[loss=0.1213, simple_loss=0.2057, pruned_loss=0.01843, over 7384.00 frames. ], tot_loss[loss=0.133, simple_loss=0.214, pruned_loss=0.02603, over 1444501.69 frames. ], batch size: 65, lr: 4.70e-03, grad_scale: 8.0 +2023-03-21 07:48:35,044 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=98078.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:48:49,249 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98106.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:48:55,917 INFO [train.py:901] (1/2) Epoch 35, batch 2100, loss[loss=0.1171, simple_loss=0.1993, pruned_loss=0.01745, over 7308.00 frames. ], tot_loss[loss=0.1328, simple_loss=0.2142, pruned_loss=0.02574, over 1445574.87 frames. ], batch size: 44, lr: 4.70e-03, grad_scale: 8.0 +2023-03-21 07:49:03,140 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7794, 1.5928, 2.0252, 2.3673, 2.0381, 2.0698, 1.9448, 2.1523], + device='cuda:1'), covar=tensor([0.5088, 0.5483, 0.3180, 0.1138, 0.5441, 0.4536, 0.3193, 0.3448], + device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0077, 0.0067, 0.0061, 0.0060, 0.0062, 0.0102, 0.0063], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:49:04,024 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 07:49:06,970 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 07:49:08,440 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.192e+02 1.595e+02 1.829e+02 2.259e+02 6.867e+02, threshold=3.658e+02, percent-clipped=3.0 +2023-03-21 07:49:09,156 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3249, 2.7953, 2.0116, 3.0498, 2.9868, 3.3348, 2.5899, 2.6028], + device='cuda:1'), covar=tensor([0.2131, 0.1104, 0.3969, 0.0545, 0.0269, 0.0391, 0.0336, 0.0398], + device='cuda:1'), in_proj_covar=tensor([0.0258, 0.0233, 0.0251, 0.0258, 0.0197, 0.0196, 0.0213, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 07:49:21,030 INFO [train.py:901] (1/2) Epoch 35, batch 2150, loss[loss=0.1142, simple_loss=0.1957, pruned_loss=0.01629, over 7320.00 frames. ], tot_loss[loss=0.1325, simple_loss=0.2138, pruned_loss=0.02561, over 1445241.27 frames. ], batch size: 44, lr: 4.70e-03, grad_scale: 8.0 +2023-03-21 07:49:21,188 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98167.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:49:34,809 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5787, 5.1203, 5.2148, 5.0840, 4.9336, 4.5387, 5.2038, 4.9849], + device='cuda:1'), covar=tensor([0.0455, 0.0365, 0.0301, 0.0465, 0.0364, 0.0431, 0.0301, 0.0442], + device='cuda:1'), in_proj_covar=tensor([0.0215, 0.0254, 0.0198, 0.0197, 0.0155, 0.0231, 0.0204, 0.0148], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:49:38,982 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 +2023-03-21 07:49:39,836 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6245, 2.8912, 3.5146, 3.6199, 3.6265, 3.6808, 3.5034, 3.4700], + device='cuda:1'), covar=tensor([0.0027, 0.0122, 0.0035, 0.0031, 0.0031, 0.0028, 0.0049, 0.0057], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0065, 0.0054, 0.0052, 0.0052, 0.0057, 0.0045, 0.0072], + device='cuda:1'), out_proj_covar=tensor([7.7262e-05, 1.3454e-04, 1.0074e-04, 9.1610e-05, 9.0409e-05, 1.0144e-04, + 8.7793e-05, 1.3479e-04], device='cuda:1') +2023-03-21 07:49:44,929 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98211.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:49:44,962 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6969, 2.4581, 3.1847, 2.8432, 2.9505, 2.7950, 2.5015, 3.0582], + device='cuda:1'), covar=tensor([0.1690, 0.0965, 0.0730, 0.1091, 0.0815, 0.1043, 0.1870, 0.1216], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0066, 0.0050, 0.0049, 0.0049, 0.0049, 0.0066, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 07:49:47,830 INFO [train.py:901] (1/2) Epoch 35, batch 2200, loss[loss=0.1315, simple_loss=0.2174, pruned_loss=0.02286, over 7215.00 frames. ], tot_loss[loss=0.1325, simple_loss=0.2136, pruned_loss=0.02566, over 1444625.82 frames. ], batch size: 50, lr: 4.70e-03, grad_scale: 8.0 +2023-03-21 07:49:50,486 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4656, 3.6181, 3.1275, 3.3462, 3.0948, 2.8683, 3.5227, 3.0481], + device='cuda:1'), covar=tensor([0.0870, 0.0556, 0.0767, 0.0932, 0.2126, 0.0722, 0.1572, 0.1779], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0057, 0.0066, 0.0058, 0.0055, 0.0059, 0.0055, 0.0054], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:49:53,855 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 07:50:00,346 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.754e+02 2.083e+02 2.530e+02 4.661e+02, threshold=4.167e+02, percent-clipped=2.0 +2023-03-21 07:50:01,393 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98244.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:50:08,896 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=98259.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:50:12,929 INFO [train.py:901] (1/2) Epoch 35, batch 2250, loss[loss=0.1193, simple_loss=0.2018, pruned_loss=0.0184, over 7353.00 frames. ], tot_loss[loss=0.1319, simple_loss=0.2132, pruned_loss=0.02531, over 1444040.50 frames. ], batch size: 63, lr: 4.70e-03, grad_scale: 8.0 +2023-03-21 07:50:15,609 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2684, 3.3253, 2.3836, 3.7413, 2.8861, 3.3597, 1.6967, 2.3903], + device='cuda:1'), covar=tensor([0.0510, 0.1097, 0.2660, 0.0661, 0.0506, 0.0797, 0.3757, 0.1883], + device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0260, 0.0284, 0.0270, 0.0273, 0.0266, 0.0238, 0.0262], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 07:50:26,739 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 07:50:27,237 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 07:50:39,648 INFO [train.py:901] (1/2) Epoch 35, batch 2300, loss[loss=0.1376, simple_loss=0.2258, pruned_loss=0.0247, over 7278.00 frames. ], tot_loss[loss=0.1323, simple_loss=0.2132, pruned_loss=0.02572, over 1441715.92 frames. ], batch size: 52, lr: 4.70e-03, grad_scale: 8.0 +2023-03-21 07:50:39,773 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4501, 4.2887, 4.0089, 4.0051, 3.6952, 2.7017, 2.3026, 4.4760], + device='cuda:1'), covar=tensor([0.0043, 0.0051, 0.0077, 0.0057, 0.0103, 0.0517, 0.0540, 0.0041], + device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0090, 0.0112, 0.0096, 0.0126, 0.0135, 0.0131, 0.0103], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 07:50:41,161 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 07:50:46,196 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98330.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 07:50:52,067 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.334e+02 1.745e+02 2.069e+02 2.499e+02 4.009e+02, threshold=4.139e+02, percent-clipped=0.0 +2023-03-21 07:50:56,751 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9692, 3.7289, 3.7061, 3.6247, 3.6108, 3.5958, 3.8779, 3.4831], + device='cuda:1'), covar=tensor([0.0135, 0.0161, 0.0115, 0.0192, 0.0434, 0.0117, 0.0136, 0.0171], + device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0100, 0.0099, 0.0089, 0.0173, 0.0106, 0.0104, 0.0111], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:51:04,687 INFO [train.py:901] (1/2) Epoch 35, batch 2350, loss[loss=0.1232, simple_loss=0.2019, pruned_loss=0.02226, over 7338.00 frames. ], tot_loss[loss=0.1329, simple_loss=0.2139, pruned_loss=0.02598, over 1443441.98 frames. ], batch size: 44, lr: 4.70e-03, grad_scale: 4.0 +2023-03-21 07:51:10,958 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=98378.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 07:51:11,969 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5333, 5.0682, 5.1710, 5.0575, 4.9225, 4.5791, 5.1766, 4.9653], + device='cuda:1'), covar=tensor([0.0411, 0.0382, 0.0339, 0.0490, 0.0322, 0.0369, 0.0291, 0.0401], + device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0252, 0.0196, 0.0195, 0.0154, 0.0228, 0.0201, 0.0147], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:51:27,843 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98410.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:51:28,226 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 07:51:31,190 INFO [train.py:901] (1/2) Epoch 35, batch 2400, loss[loss=0.1307, simple_loss=0.2119, pruned_loss=0.02477, over 7291.00 frames. ], tot_loss[loss=0.1332, simple_loss=0.2139, pruned_loss=0.02624, over 1443533.05 frames. ], batch size: 66, lr: 4.70e-03, grad_scale: 8.0 +2023-03-21 07:51:33,274 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.2336, 5.6660, 5.7149, 5.6562, 5.4460, 5.3009, 5.7820, 5.5862], + device='cuda:1'), covar=tensor([0.0385, 0.0320, 0.0332, 0.0405, 0.0269, 0.0264, 0.0266, 0.0350], + device='cuda:1'), in_proj_covar=tensor([0.0213, 0.0253, 0.0197, 0.0195, 0.0155, 0.0229, 0.0202, 0.0147], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:51:34,752 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 07:51:44,316 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.114e+02 1.798e+02 2.138e+02 2.531e+02 4.727e+02, threshold=4.276e+02, percent-clipped=2.0 +2023-03-21 07:51:44,868 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 07:51:47,364 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 07:51:49,978 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0330, 2.7394, 3.1673, 3.0636, 3.0584, 2.9209, 2.6387, 3.1889], + device='cuda:1'), covar=tensor([0.1287, 0.0849, 0.1091, 0.1093, 0.0919, 0.1261, 0.1542, 0.1177], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0065, 0.0049, 0.0048, 0.0047, 0.0048, 0.0065, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 07:51:54,491 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98462.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:51:57,513 INFO [train.py:901] (1/2) Epoch 35, batch 2450, loss[loss=0.1085, simple_loss=0.1842, pruned_loss=0.01644, over 6919.00 frames. ], tot_loss[loss=0.1332, simple_loss=0.2138, pruned_loss=0.02628, over 1441777.79 frames. ], batch size: 35, lr: 4.69e-03, grad_scale: 8.0 +2023-03-21 07:51:59,635 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98471.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:52:14,147 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 07:52:22,566 INFO [train.py:901] (1/2) Epoch 35, batch 2500, loss[loss=0.1322, simple_loss=0.2178, pruned_loss=0.02327, over 7301.00 frames. ], tot_loss[loss=0.1336, simple_loss=0.2141, pruned_loss=0.02654, over 1442535.77 frames. ], batch size: 86, lr: 4.69e-03, grad_scale: 8.0 +2023-03-21 07:52:31,479 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2701, 2.4498, 2.6335, 2.5030, 2.4555, 2.4829, 2.3690, 2.1488], + device='cuda:1'), covar=tensor([0.0455, 0.0514, 0.0434, 0.0308, 0.0424, 0.0494, 0.0245, 0.0316], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0036, 0.0038, 0.0036, 0.0035, 0.0035, 0.0040, 0.0039], + device='cuda:1'), out_proj_covar=tensor([9.6301e-05, 9.4736e-05, 9.5142e-05, 9.1070e-05, 9.1223e-05, 9.0005e-05, + 9.7886e-05, 9.9248e-05], device='cuda:1') +2023-03-21 07:52:35,897 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.356e+02 1.755e+02 2.080e+02 2.550e+02 4.221e+02, threshold=4.161e+02, percent-clipped=0.0 +2023-03-21 07:52:36,074 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98543.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:52:36,506 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98544.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:52:40,603 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 07:52:45,855 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.8861, 5.4380, 5.5177, 5.4580, 5.2300, 4.9948, 5.5651, 5.3698], + device='cuda:1'), covar=tensor([0.0458, 0.0356, 0.0427, 0.0459, 0.0277, 0.0347, 0.0304, 0.0373], + device='cuda:1'), in_proj_covar=tensor([0.0213, 0.0253, 0.0198, 0.0195, 0.0154, 0.0229, 0.0202, 0.0149], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:52:49,280 INFO [train.py:901] (1/2) Epoch 35, batch 2550, loss[loss=0.1381, simple_loss=0.2221, pruned_loss=0.02711, over 7288.00 frames. ], tot_loss[loss=0.1328, simple_loss=0.2136, pruned_loss=0.02604, over 1443141.60 frames. ], batch size: 57, lr: 4.69e-03, grad_scale: 8.0 +2023-03-21 07:52:58,880 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0328, 2.5845, 1.9756, 2.9728, 2.7878, 3.0666, 2.6427, 2.5434], + device='cuda:1'), covar=tensor([0.2397, 0.1115, 0.3740, 0.0596, 0.0298, 0.0268, 0.0370, 0.0351], + device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0233, 0.0252, 0.0256, 0.0196, 0.0196, 0.0214, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 07:53:01,800 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=98592.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:53:06,284 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1575, 3.9713, 3.3690, 3.6746, 3.0712, 2.4604, 1.9870, 4.2076], + device='cuda:1'), covar=tensor([0.0057, 0.0086, 0.0156, 0.0093, 0.0188, 0.0559, 0.0625, 0.0052], + device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0090, 0.0112, 0.0096, 0.0126, 0.0134, 0.0131, 0.0103], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 07:53:08,344 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98604.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:53:14,788 INFO [train.py:901] (1/2) Epoch 35, batch 2600, loss[loss=0.1453, simple_loss=0.2218, pruned_loss=0.03438, over 7355.00 frames. ], tot_loss[loss=0.1331, simple_loss=0.2136, pruned_loss=0.02631, over 1442924.76 frames. ], batch size: 73, lr: 4.69e-03, grad_scale: 8.0 +2023-03-21 07:53:18,864 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1812, 3.8707, 3.8205, 3.8326, 3.8467, 3.7427, 4.0147, 3.5984], + device='cuda:1'), covar=tensor([0.0152, 0.0187, 0.0146, 0.0204, 0.0472, 0.0146, 0.0183, 0.0195], + device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0101, 0.0100, 0.0090, 0.0175, 0.0108, 0.0105, 0.0112], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:53:26,461 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 +2023-03-21 07:53:27,538 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.227e+02 1.759e+02 1.926e+02 2.182e+02 4.869e+02, threshold=3.852e+02, percent-clipped=2.0 +2023-03-21 07:53:39,557 INFO [train.py:901] (1/2) Epoch 35, batch 2650, loss[loss=0.1404, simple_loss=0.2208, pruned_loss=0.02999, over 7276.00 frames. ], tot_loss[loss=0.1332, simple_loss=0.2138, pruned_loss=0.02631, over 1443074.53 frames. ], batch size: 52, lr: 4.69e-03, grad_scale: 8.0 +2023-03-21 07:54:05,467 INFO [train.py:901] (1/2) Epoch 35, batch 2700, loss[loss=0.1394, simple_loss=0.2134, pruned_loss=0.03271, over 7319.00 frames. ], tot_loss[loss=0.1329, simple_loss=0.2135, pruned_loss=0.02617, over 1442116.28 frames. ], batch size: 49, lr: 4.69e-03, grad_scale: 8.0 +2023-03-21 07:54:18,203 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.314e+02 1.808e+02 2.115e+02 2.423e+02 3.699e+02, threshold=4.229e+02, percent-clipped=0.0 +2023-03-21 07:54:27,529 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98762.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:54:29,406 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98766.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:54:29,837 INFO [train.py:901] (1/2) Epoch 35, batch 2750, loss[loss=0.1388, simple_loss=0.2167, pruned_loss=0.03046, over 7267.00 frames. ], tot_loss[loss=0.1334, simple_loss=0.2144, pruned_loss=0.02621, over 1444070.32 frames. ], batch size: 52, lr: 4.69e-03, grad_scale: 8.0 +2023-03-21 07:54:37,157 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 +2023-03-21 07:54:51,380 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=98810.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:54:54,673 INFO [train.py:901] (1/2) Epoch 35, batch 2800, loss[loss=0.1488, simple_loss=0.2339, pruned_loss=0.03182, over 7263.00 frames. ], tot_loss[loss=0.1334, simple_loss=0.2143, pruned_loss=0.02626, over 1444437.89 frames. ], batch size: 89, lr: 4.69e-03, grad_scale: 8.0 +2023-03-21 07:54:55,070 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 +2023-03-21 07:54:57,409 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-03-21 07:55:20,562 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 07:55:20,615 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 07:55:20,640 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 07:55:20,657 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 07:55:20,827 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 07:55:20,841 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 07:55:20,853 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 07:55:20,949 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 07:55:21,343 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 07:55:21,358 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 07:55:21,375 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 07:55:21,538 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 07:55:21,544 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 07:55:21,551 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 07:55:21,688 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 07:55:21,691 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 07:55:21,751 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 07:55:26,022 INFO [train.py:901] (1/2) Epoch 36, batch 0, loss[loss=0.1326, simple_loss=0.2169, pruned_loss=0.02418, over 7290.00 frames. ], tot_loss[loss=0.1326, simple_loss=0.2169, pruned_loss=0.02418, over 7290.00 frames. ], batch size: 70, lr: 4.62e-03, grad_scale: 8.0 +2023-03-21 07:55:26,022 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 07:55:32,495 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.1131, 5.3105, 5.3919, 5.3536, 4.9763, 4.9193, 5.3676, 5.0311], + device='cuda:1'), covar=tensor([0.0278, 0.0308, 0.0308, 0.0390, 0.0341, 0.0286, 0.0287, 0.0469], + device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0251, 0.0196, 0.0195, 0.0154, 0.0229, 0.0203, 0.0148], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:55:39,716 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4869, 4.0887, 4.0292, 4.1225, 4.1044, 4.0386, 4.1570, 3.8384], + device='cuda:1'), covar=tensor([0.0099, 0.0173, 0.0134, 0.0177, 0.0478, 0.0121, 0.0188, 0.0206], + device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0101, 0.0100, 0.0090, 0.0175, 0.0107, 0.0105, 0.0111], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 07:55:51,379 INFO [train.py:935] (1/2) Epoch 36, validation: loss=0.1652, simple_loss=0.2566, pruned_loss=0.03687, over 1622729.00 frames. +2023-03-21 07:55:51,380 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 07:55:52,367 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.696e+02 1.971e+02 2.312e+02 4.526e+02, threshold=3.942e+02, percent-clipped=1.0 +2023-03-21 07:55:58,411 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 07:56:08,345 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 07:56:16,039 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 07:56:16,645 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98890.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:56:16,992 INFO [train.py:901] (1/2) Epoch 36, batch 50, loss[loss=0.1237, simple_loss=0.2141, pruned_loss=0.01665, over 7325.00 frames. ], tot_loss[loss=0.1347, simple_loss=0.215, pruned_loss=0.02722, over 325251.00 frames. ], batch size: 75, lr: 4.62e-03, grad_scale: 8.0 +2023-03-21 07:56:18,048 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 07:56:19,851 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98895.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:56:21,215 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 07:56:21,788 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98899.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:56:39,122 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 07:56:39,137 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 07:56:42,649 INFO [train.py:901] (1/2) Epoch 36, batch 100, loss[loss=0.13, simple_loss=0.2109, pruned_loss=0.02455, over 7291.00 frames. ], tot_loss[loss=0.1334, simple_loss=0.2142, pruned_loss=0.02633, over 573432.48 frames. ], batch size: 68, lr: 4.62e-03, grad_scale: 8.0 +2023-03-21 07:56:43,623 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.846e+02 2.136e+02 2.564e+02 5.371e+02, threshold=4.271e+02, percent-clipped=2.0 +2023-03-21 07:56:47,824 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98951.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:56:50,440 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98956.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:56:52,152 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 +2023-03-21 07:56:57,849 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98971.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:57:08,775 INFO [train.py:901] (1/2) Epoch 36, batch 150, loss[loss=0.1438, simple_loss=0.223, pruned_loss=0.03224, over 7297.00 frames. ], tot_loss[loss=0.133, simple_loss=0.2141, pruned_loss=0.02597, over 766080.92 frames. ], batch size: 68, lr: 4.62e-03, grad_scale: 8.0 +2023-03-21 07:57:29,882 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99032.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:57:30,499 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 +2023-03-21 07:57:34,200 INFO [train.py:901] (1/2) Epoch 36, batch 200, loss[loss=0.1534, simple_loss=0.2283, pruned_loss=0.03922, over 7243.00 frames. ], tot_loss[loss=0.1333, simple_loss=0.2146, pruned_loss=0.02601, over 918070.40 frames. ], batch size: 55, lr: 4.61e-03, grad_scale: 8.0 +2023-03-21 07:57:35,163 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.253e+02 1.668e+02 1.975e+02 2.351e+02 4.108e+02, threshold=3.950e+02, percent-clipped=0.0 +2023-03-21 07:57:38,696 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 07:57:39,274 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99051.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:57:42,628 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 07:57:47,906 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99066.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:57:50,366 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 07:57:56,021 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9536, 1.7380, 2.2649, 2.5276, 2.3520, 2.4097, 2.4015, 2.5045], + device='cuda:1'), covar=tensor([0.2186, 0.4692, 0.1869, 0.1817, 0.3161, 0.3240, 0.2618, 0.1767], + device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0078, 0.0068, 0.0064, 0.0062, 0.0062, 0.0104, 0.0065], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:58:00,459 INFO [train.py:901] (1/2) Epoch 36, batch 250, loss[loss=0.1313, simple_loss=0.216, pruned_loss=0.02324, over 7301.00 frames. ], tot_loss[loss=0.132, simple_loss=0.2129, pruned_loss=0.02551, over 1033461.83 frames. ], batch size: 80, lr: 4.61e-03, grad_scale: 8.0 +2023-03-21 07:58:02,947 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 07:58:10,936 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99112.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:58:11,851 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=99114.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:58:17,181 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.86 vs. limit=5.0 +2023-03-21 07:58:23,877 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 07:58:25,399 INFO [train.py:901] (1/2) Epoch 36, batch 300, loss[loss=0.1193, simple_loss=0.1949, pruned_loss=0.0219, over 7311.00 frames. ], tot_loss[loss=0.1314, simple_loss=0.2125, pruned_loss=0.02518, over 1123824.28 frames. ], batch size: 42, lr: 4.61e-03, grad_scale: 4.0 +2023-03-21 07:58:26,855 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.227e+02 1.659e+02 1.981e+02 2.270e+02 3.464e+02, threshold=3.962e+02, percent-clipped=0.0 +2023-03-21 07:58:33,801 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 +2023-03-21 07:58:33,869 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 07:58:51,381 INFO [train.py:901] (1/2) Epoch 36, batch 350, loss[loss=0.1446, simple_loss=0.2144, pruned_loss=0.03737, over 7328.00 frames. ], tot_loss[loss=0.1314, simple_loss=0.2122, pruned_loss=0.02528, over 1194326.94 frames. ], batch size: 49, lr: 4.61e-03, grad_scale: 4.0 +2023-03-21 07:58:55,470 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99199.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:58:57,028 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-21 07:59:06,566 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 07:59:10,613 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99229.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:59:11,694 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4282, 3.0801, 3.5323, 3.4119, 3.0662, 2.9649, 3.5764, 2.6910], + device='cuda:1'), covar=tensor([0.0379, 0.0550, 0.0614, 0.0538, 0.0617, 0.0861, 0.0518, 0.1895], + device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0336, 0.0272, 0.0352, 0.0290, 0.0290, 0.0343, 0.0254], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:59:17,815 INFO [train.py:901] (1/2) Epoch 36, batch 400, loss[loss=0.1161, simple_loss=0.2024, pruned_loss=0.0149, over 7318.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.2128, pruned_loss=0.02538, over 1249810.66 frames. ], batch size: 75, lr: 4.61e-03, grad_scale: 8.0 +2023-03-21 07:59:19,285 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.276e+02 1.771e+02 2.180e+02 2.603e+02 7.277e+02, threshold=4.360e+02, percent-clipped=4.0 +2023-03-21 07:59:20,372 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99246.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:59:20,835 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=99247.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:59:22,867 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99251.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:59:30,652 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 +2023-03-21 07:59:30,873 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3158, 4.7871, 4.8379, 4.8587, 4.7390, 4.3794, 4.8763, 4.7420], + device='cuda:1'), covar=tensor([0.0462, 0.0414, 0.0394, 0.0404, 0.0340, 0.0388, 0.0359, 0.0439], + device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0253, 0.0198, 0.0195, 0.0155, 0.0228, 0.0204, 0.0149], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 07:59:31,442 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99268.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:59:42,546 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99290.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 07:59:42,903 INFO [train.py:901] (1/2) Epoch 36, batch 450, loss[loss=0.1127, simple_loss=0.1855, pruned_loss=0.01995, over 7005.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.2128, pruned_loss=0.02537, over 1292318.46 frames. ], batch size: 35, lr: 4.61e-03, grad_scale: 4.0 +2023-03-21 07:59:44,069 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0307, 3.1322, 3.3496, 3.0595, 3.2756, 3.2544, 2.7138, 3.2781], + device='cuda:1'), covar=tensor([0.1662, 0.0592, 0.1075, 0.1228, 0.1284, 0.0842, 0.2239, 0.1296], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0066, 0.0051, 0.0049, 0.0049, 0.0048, 0.0067, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 07:59:46,960 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 07:59:47,493 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 07:59:59,094 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.13 vs. limit=5.0 +2023-03-21 08:00:01,059 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99324.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:00:02,511 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99327.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:00:03,609 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99329.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:00:09,418 INFO [train.py:901] (1/2) Epoch 36, batch 500, loss[loss=0.1314, simple_loss=0.2226, pruned_loss=0.02009, over 7215.00 frames. ], tot_loss[loss=0.1315, simple_loss=0.2122, pruned_loss=0.02541, over 1323028.50 frames. ], batch size: 93, lr: 4.61e-03, grad_scale: 4.0 +2023-03-21 08:00:11,403 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.249e+02 1.683e+02 2.019e+02 2.214e+02 6.432e+02, threshold=4.039e+02, percent-clipped=1.0 +2023-03-21 08:00:21,411 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 08:00:22,394 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 08:00:23,434 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 08:00:25,020 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 08:00:30,051 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 08:00:31,632 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99385.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:00:34,388 INFO [train.py:901] (1/2) Epoch 36, batch 550, loss[loss=0.1322, simple_loss=0.2019, pruned_loss=0.03127, over 7221.00 frames. ], tot_loss[loss=0.1312, simple_loss=0.2118, pruned_loss=0.02531, over 1348605.46 frames. ], batch size: 45, lr: 4.61e-03, grad_scale: 4.0 +2023-03-21 08:00:38,668 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 +2023-03-21 08:00:41,359 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 08:00:43,460 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99407.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:00:45,646 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0868, 3.5928, 2.7389, 3.1069, 3.2939, 3.0443, 3.4409, 2.9138], + device='cuda:1'), covar=tensor([0.0648, 0.0631, 0.1068, 0.1103, 0.1519, 0.0784, 0.0812, 0.0855], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0057, 0.0066, 0.0058, 0.0055, 0.0059, 0.0056, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:00:49,540 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 08:00:49,917 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 +2023-03-21 08:00:50,675 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99420.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:00:53,111 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 08:00:59,607 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 08:01:00,743 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2641, 4.0846, 3.6161, 3.8214, 3.1736, 2.6062, 2.1552, 4.2558], + device='cuda:1'), covar=tensor([0.0047, 0.0061, 0.0115, 0.0068, 0.0159, 0.0489, 0.0573, 0.0044], + device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0091, 0.0112, 0.0094, 0.0126, 0.0134, 0.0130, 0.0102], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 08:01:01,126 INFO [train.py:901] (1/2) Epoch 36, batch 600, loss[loss=0.1432, simple_loss=0.2215, pruned_loss=0.03242, over 7340.00 frames. ], tot_loss[loss=0.1311, simple_loss=0.2117, pruned_loss=0.02527, over 1368594.52 frames. ], batch size: 54, lr: 4.61e-03, grad_scale: 4.0 +2023-03-21 08:01:03,078 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.742e+02 2.120e+02 2.350e+02 3.702e+02, threshold=4.240e+02, percent-clipped=0.0 +2023-03-21 08:01:14,738 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99468.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 08:01:15,667 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 08:01:21,396 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99481.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:01:23,815 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 08:01:27,481 INFO [train.py:901] (1/2) Epoch 36, batch 650, loss[loss=0.141, simple_loss=0.2241, pruned_loss=0.02893, over 7349.00 frames. ], tot_loss[loss=0.1316, simple_loss=0.2124, pruned_loss=0.02538, over 1384714.04 frames. ], batch size: 63, lr: 4.60e-03, grad_scale: 4.0 +2023-03-21 08:01:41,990 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 08:01:46,655 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99529.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 08:01:51,007 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 08:01:51,174 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6452, 3.5959, 2.6569, 4.0103, 3.5273, 3.5822, 1.7410, 2.6076], + device='cuda:1'), covar=tensor([0.0583, 0.0622, 0.2780, 0.0480, 0.0484, 0.0715, 0.3966, 0.1959], + device='cuda:1'), in_proj_covar=tensor([0.0284, 0.0261, 0.0289, 0.0273, 0.0273, 0.0269, 0.0241, 0.0265], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:01:52,517 INFO [train.py:901] (1/2) Epoch 36, batch 700, loss[loss=0.1408, simple_loss=0.2196, pruned_loss=0.03099, over 7350.00 frames. ], tot_loss[loss=0.1312, simple_loss=0.2118, pruned_loss=0.02526, over 1394178.28 frames. ], batch size: 51, lr: 4.60e-03, grad_scale: 4.0 +2023-03-21 08:01:54,521 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.217e+02 1.727e+02 1.965e+02 2.188e+02 5.872e+02, threshold=3.931e+02, percent-clipped=1.0 +2023-03-21 08:01:55,074 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99546.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:01:57,603 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99551.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:02:06,060 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99568.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:02:15,915 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99585.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:02:16,357 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 08:02:17,355 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 08:02:18,006 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99589.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:02:18,839 INFO [train.py:901] (1/2) Epoch 36, batch 750, loss[loss=0.1153, simple_loss=0.1945, pruned_loss=0.01806, over 7260.00 frames. ], tot_loss[loss=0.1315, simple_loss=0.2119, pruned_loss=0.02555, over 1405040.84 frames. ], batch size: 47, lr: 4.60e-03, grad_scale: 4.0 +2023-03-21 08:02:20,370 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=99594.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:02:20,453 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8692, 3.4779, 3.6791, 3.3736, 3.5891, 3.3368, 2.9059, 3.2636], + device='cuda:1'), covar=tensor([0.2418, 0.0674, 0.0845, 0.1194, 0.0787, 0.1035, 0.2454, 0.1577], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0066, 0.0051, 0.0050, 0.0050, 0.0048, 0.0067, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 08:02:22,871 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=99599.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:02:31,194 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 08:02:35,740 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 08:02:35,793 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99624.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:02:37,358 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99627.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:02:38,460 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99629.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:02:41,828 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 08:02:42,802 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 08:02:43,154 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 +2023-03-21 08:02:44,290 INFO [train.py:901] (1/2) Epoch 36, batch 800, loss[loss=0.1376, simple_loss=0.2257, pruned_loss=0.02482, over 7281.00 frames. ], tot_loss[loss=0.1317, simple_loss=0.2125, pruned_loss=0.0255, over 1415507.24 frames. ], batch size: 68, lr: 4.60e-03, grad_scale: 8.0 +2023-03-21 08:02:46,282 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.255e+02 1.802e+02 2.004e+02 2.443e+02 3.511e+02, threshold=4.007e+02, percent-clipped=0.0 +2023-03-21 08:02:49,001 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99650.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:02:53,843 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 08:03:02,634 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=99675.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:03:05,141 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:03:05,180 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:03:08,408 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 +2023-03-21 08:03:10,531 INFO [train.py:901] (1/2) Epoch 36, batch 850, loss[loss=0.1122, simple_loss=0.1883, pruned_loss=0.01805, over 6980.00 frames. ], tot_loss[loss=0.1323, simple_loss=0.2131, pruned_loss=0.02575, over 1421852.77 frames. ], batch size: 35, lr: 4.60e-03, grad_scale: 8.0 +2023-03-21 08:03:12,975 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 08:03:13,419 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 08:03:18,350 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 08:03:18,459 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99707.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:03:21,905 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 08:03:24,623 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7211, 2.4700, 2.4028, 3.7199, 1.8136, 3.4832, 1.5716, 3.1831], + device='cuda:1'), covar=tensor([0.0161, 0.1472, 0.1821, 0.0201, 0.4092, 0.0291, 0.1290, 0.0302], + device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0252, 0.0265, 0.0208, 0.0251, 0.0213, 0.0231, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:03:35,384 INFO [train.py:901] (1/2) Epoch 36, batch 900, loss[loss=0.1421, simple_loss=0.2211, pruned_loss=0.03154, over 7240.00 frames. ], tot_loss[loss=0.1321, simple_loss=0.2132, pruned_loss=0.02553, over 1427095.47 frames. ], batch size: 55, lr: 4.60e-03, grad_scale: 8.0 +2023-03-21 08:03:35,536 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99741.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:03:37,377 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.709e+02 1.934e+02 2.363e+02 3.874e+02, threshold=3.867e+02, percent-clipped=0.0 +2023-03-21 08:03:43,190 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=99755.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:03:52,878 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9912, 4.1740, 4.0075, 4.1794, 3.8742, 3.9476, 4.3364, 4.4015], + device='cuda:1'), covar=tensor([0.0311, 0.0200, 0.0258, 0.0248, 0.0375, 0.0396, 0.0302, 0.0243], + device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0128, 0.0119, 0.0125, 0.0115, 0.0106, 0.0099, 0.0101], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:03:54,357 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99776.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:03:59,861 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 08:04:00,370 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5005, 5.0287, 5.1167, 5.0170, 4.8827, 4.5604, 5.1106, 4.9077], + device='cuda:1'), covar=tensor([0.0457, 0.0391, 0.0341, 0.0484, 0.0348, 0.0430, 0.0335, 0.0454], + device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0252, 0.0196, 0.0196, 0.0154, 0.0228, 0.0204, 0.0148], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 08:04:01,796 INFO [train.py:901] (1/2) Epoch 36, batch 950, loss[loss=0.1173, simple_loss=0.198, pruned_loss=0.01832, over 7303.00 frames. ], tot_loss[loss=0.1322, simple_loss=0.2133, pruned_loss=0.02553, over 1432772.28 frames. ], batch size: 83, lr: 4.60e-03, grad_scale: 8.0 +2023-03-21 08:04:18,754 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99824.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 08:04:24,229 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 08:04:25,311 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7714, 3.0950, 3.5684, 3.6636, 3.7782, 3.8354, 3.7258, 3.6322], + device='cuda:1'), covar=tensor([0.0029, 0.0121, 0.0040, 0.0037, 0.0034, 0.0029, 0.0047, 0.0051], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0068, 0.0056, 0.0055, 0.0054, 0.0060, 0.0047, 0.0075], + device='cuda:1'), out_proj_covar=tensor([8.1368e-05, 1.3989e-04, 1.0319e-04, 9.8070e-05, 9.3757e-05, 1.0628e-04, + 9.1960e-05, 1.4037e-04], device='cuda:1') +2023-03-21 08:04:27,781 INFO [train.py:901] (1/2) Epoch 36, batch 1000, loss[loss=0.1328, simple_loss=0.2097, pruned_loss=0.02799, over 7277.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.2129, pruned_loss=0.02534, over 1437233.10 frames. ], batch size: 52, lr: 4.60e-03, grad_scale: 8.0 +2023-03-21 08:04:29,777 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.174e+02 1.667e+02 1.934e+02 2.174e+02 4.994e+02, threshold=3.868e+02, percent-clipped=1.0 +2023-03-21 08:04:34,188 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-03-21 08:04:45,788 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 08:04:50,291 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99885.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:04:53,207 INFO [train.py:901] (1/2) Epoch 36, batch 1050, loss[loss=0.1209, simple_loss=0.2076, pruned_loss=0.01717, over 7344.00 frames. ], tot_loss[loss=0.1321, simple_loss=0.2131, pruned_loss=0.02556, over 1439084.53 frames. ], batch size: 44, lr: 4.60e-03, grad_scale: 8.0 +2023-03-21 08:05:06,194 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 08:05:09,835 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:05:09,886 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:05:10,305 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 08:05:15,025 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=99933.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:05:19,603 INFO [train.py:901] (1/2) Epoch 36, batch 1100, loss[loss=0.1501, simple_loss=0.2313, pruned_loss=0.03444, over 7302.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.2128, pruned_loss=0.02535, over 1440359.62 frames. ], batch size: 83, lr: 4.59e-03, grad_scale: 8.0 +2023-03-21 08:05:21,598 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.278e+02 1.809e+02 2.088e+02 2.567e+02 3.925e+02, threshold=4.176e+02, percent-clipped=1.0 +2023-03-21 08:05:21,690 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99945.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:05:25,593 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 +2023-03-21 08:05:35,338 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=99972.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:05:39,411 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99980.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:05:39,802 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 08:05:40,318 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 08:05:44,763 INFO [train.py:901] (1/2) Epoch 36, batch 1150, loss[loss=0.1176, simple_loss=0.1882, pruned_loss=0.02347, over 7143.00 frames. ], tot_loss[loss=0.1315, simple_loss=0.2123, pruned_loss=0.02536, over 1439088.15 frames. ], batch size: 39, lr: 4.59e-03, grad_scale: 8.0 +2023-03-21 08:05:54,222 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0979, 2.4732, 1.8668, 2.8220, 2.8079, 2.6470, 2.5401, 2.4579], + device='cuda:1'), covar=tensor([0.1989, 0.1025, 0.3528, 0.0748, 0.0295, 0.0227, 0.0359, 0.0347], + device='cuda:1'), in_proj_covar=tensor([0.0254, 0.0232, 0.0249, 0.0260, 0.0195, 0.0198, 0.0214, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 08:05:56,602 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 08:05:56,735 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100007.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 08:05:57,128 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 08:06:08,470 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100028.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:06:12,491 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100036.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:06:14,936 INFO [train.py:901] (1/2) Epoch 36, batch 1200, loss[loss=0.1497, simple_loss=0.2275, pruned_loss=0.03593, over 7274.00 frames. ], tot_loss[loss=0.131, simple_loss=0.212, pruned_loss=0.02501, over 1441909.48 frames. ], batch size: 66, lr: 4.59e-03, grad_scale: 8.0 +2023-03-21 08:06:16,956 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.316e+02 1.656e+02 1.985e+02 2.321e+02 4.535e+02, threshold=3.970e+02, percent-clipped=2.0 +2023-03-21 08:06:22,799 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-21 08:06:25,632 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1171, 2.3392, 2.4997, 2.1320, 2.3092, 2.1951, 2.0005, 1.7698], + device='cuda:1'), covar=tensor([0.0467, 0.0443, 0.0164, 0.0204, 0.0509, 0.0444, 0.0366, 0.0330], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0037, 0.0038, 0.0037, 0.0035, 0.0035, 0.0041, 0.0040], + device='cuda:1'), out_proj_covar=tensor([9.7242e-05, 9.6482e-05, 9.5646e-05, 9.2979e-05, 9.1695e-05, 9.1785e-05, + 1.0071e-04, 1.0069e-04], device='cuda:1') +2023-03-21 08:06:28,676 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100068.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 08:06:29,515 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 08:06:32,129 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0444, 3.9750, 3.2591, 3.6417, 3.0931, 2.5871, 1.8107, 4.1001], + device='cuda:1'), covar=tensor([0.0054, 0.0067, 0.0145, 0.0079, 0.0160, 0.0462, 0.0681, 0.0045], + device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0091, 0.0112, 0.0094, 0.0126, 0.0133, 0.0130, 0.0102], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 08:06:32,617 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100076.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:06:40,125 INFO [train.py:901] (1/2) Epoch 36, batch 1250, loss[loss=0.1188, simple_loss=0.1972, pruned_loss=0.02016, over 7227.00 frames. ], tot_loss[loss=0.1316, simple_loss=0.2128, pruned_loss=0.02518, over 1442605.05 frames. ], batch size: 45, lr: 4.59e-03, grad_scale: 8.0 +2023-03-21 08:06:52,834 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 08:06:57,091 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-03-21 08:06:57,345 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 08:06:57,401 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100124.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:06:57,444 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100124.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 08:06:57,690 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-03-21 08:06:58,396 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 08:06:59,800 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.12 vs. limit=2.0 +2023-03-21 08:07:06,033 INFO [train.py:901] (1/2) Epoch 36, batch 1300, loss[loss=0.122, simple_loss=0.1921, pruned_loss=0.02594, over 7051.00 frames. ], tot_loss[loss=0.1309, simple_loss=0.212, pruned_loss=0.02494, over 1440203.69 frames. ], batch size: 35, lr: 4.59e-03, grad_scale: 8.0 +2023-03-21 08:07:08,001 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 1.807e+02 2.023e+02 2.519e+02 3.983e+02, threshold=4.046e+02, percent-clipped=1.0 +2023-03-21 08:07:21,896 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100172.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 08:07:22,816 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 08:07:25,261 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 08:07:28,704 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 08:07:31,167 INFO [train.py:901] (1/2) Epoch 36, batch 1350, loss[loss=0.1376, simple_loss=0.2221, pruned_loss=0.02661, over 7258.00 frames. ], tot_loss[loss=0.1319, simple_loss=0.2127, pruned_loss=0.02554, over 1441232.75 frames. ], batch size: 55, lr: 4.59e-03, grad_scale: 8.0 +2023-03-21 08:07:39,700 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 08:07:41,111 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 +2023-03-21 08:07:43,773 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.1998, 4.7932, 4.6492, 5.2255, 5.0245, 5.1304, 4.4080, 4.7648], + device='cuda:1'), covar=tensor([0.0752, 0.1971, 0.2122, 0.0758, 0.0935, 0.0983, 0.0772, 0.0983], + device='cuda:1'), in_proj_covar=tensor([0.0254, 0.0378, 0.0290, 0.0299, 0.0226, 0.0357, 0.0221, 0.0263], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:07:49,308 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100224.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:07:53,084 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 +2023-03-21 08:07:56,340 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100238.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:07:57,702 INFO [train.py:901] (1/2) Epoch 36, batch 1400, loss[loss=0.1377, simple_loss=0.2113, pruned_loss=0.03201, over 7283.00 frames. ], tot_loss[loss=0.1323, simple_loss=0.2132, pruned_loss=0.0257, over 1442587.36 frames. ], batch size: 70, lr: 4.59e-03, grad_scale: 8.0 +2023-03-21 08:07:59,673 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.298e+02 1.761e+02 2.095e+02 2.393e+02 3.680e+02, threshold=4.190e+02, percent-clipped=0.0 +2023-03-21 08:07:59,830 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100245.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:08:10,715 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 08:08:13,281 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100272.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:08:17,954 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2199, 2.4067, 2.5022, 2.2305, 2.3439, 2.2635, 2.1719, 1.9075], + device='cuda:1'), covar=tensor([0.0429, 0.0406, 0.0203, 0.0283, 0.0553, 0.0523, 0.0362, 0.0353], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0037, 0.0038, 0.0037, 0.0035, 0.0036, 0.0040, 0.0040], + device='cuda:1'), out_proj_covar=tensor([9.7036e-05, 9.6913e-05, 9.5963e-05, 9.3076e-05, 9.2196e-05, 9.2145e-05, + 9.9902e-05, 1.0032e-04], device='cuda:1') +2023-03-21 08:08:23,913 INFO [train.py:901] (1/2) Epoch 36, batch 1450, loss[loss=0.1252, simple_loss=0.21, pruned_loss=0.02025, over 7315.00 frames. ], tot_loss[loss=0.1321, simple_loss=0.2131, pruned_loss=0.02552, over 1444105.78 frames. ], batch size: 80, lr: 4.59e-03, grad_scale: 8.0 +2023-03-21 08:08:25,029 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100293.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:08:28,173 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100299.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:08:34,099 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 08:08:46,820 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100336.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:08:48,523 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 +2023-03-21 08:08:49,199 INFO [train.py:901] (1/2) Epoch 36, batch 1500, loss[loss=0.09031, simple_loss=0.1589, pruned_loss=0.01087, over 5846.00 frames. ], tot_loss[loss=0.1314, simple_loss=0.2125, pruned_loss=0.02518, over 1439209.16 frames. ], batch size: 25, lr: 4.59e-03, grad_scale: 8.0 +2023-03-21 08:08:50,193 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 08:08:51,158 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.275e+02 1.764e+02 1.940e+02 2.308e+02 3.586e+02, threshold=3.880e+02, percent-clipped=0.0 +2023-03-21 08:09:00,332 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100363.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 08:09:12,000 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100384.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:09:15,496 INFO [train.py:901] (1/2) Epoch 36, batch 1550, loss[loss=0.1188, simple_loss=0.2042, pruned_loss=0.0167, over 7293.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.213, pruned_loss=0.0253, over 1440449.31 frames. ], batch size: 86, lr: 4.58e-03, grad_scale: 8.0 +2023-03-21 08:09:15,505 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 08:09:21,608 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2622, 4.0981, 3.6759, 3.7214, 3.0066, 2.4061, 1.9629, 4.2298], + device='cuda:1'), covar=tensor([0.0041, 0.0057, 0.0095, 0.0076, 0.0170, 0.0527, 0.0626, 0.0043], + device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0091, 0.0113, 0.0095, 0.0128, 0.0135, 0.0131, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 08:09:21,630 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100402.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:09:41,388 INFO [train.py:901] (1/2) Epoch 36, batch 1600, loss[loss=0.132, simple_loss=0.212, pruned_loss=0.02593, over 7296.00 frames. ], tot_loss[loss=0.1319, simple_loss=0.2133, pruned_loss=0.02524, over 1441299.68 frames. ], batch size: 49, lr: 4.58e-03, grad_scale: 8.0 +2023-03-21 08:09:43,323 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.302e+02 1.792e+02 2.169e+02 2.475e+02 4.833e+02, threshold=4.339e+02, percent-clipped=2.0 +2023-03-21 08:09:46,610 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 08:09:47,634 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 08:09:51,261 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 08:09:53,920 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100463.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 08:10:00,433 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 08:10:04,401 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 08:10:07,878 INFO [train.py:901] (1/2) Epoch 36, batch 1650, loss[loss=0.1535, simple_loss=0.2334, pruned_loss=0.0368, over 7136.00 frames. ], tot_loss[loss=0.1321, simple_loss=0.2134, pruned_loss=0.02543, over 1441258.17 frames. ], batch size: 98, lr: 4.58e-03, grad_scale: 8.0 +2023-03-21 08:10:12,396 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 08:10:30,737 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 08:10:33,845 INFO [train.py:901] (1/2) Epoch 36, batch 1700, loss[loss=0.1053, simple_loss=0.175, pruned_loss=0.01784, over 6387.00 frames. ], tot_loss[loss=0.1321, simple_loss=0.2133, pruned_loss=0.02545, over 1441508.19 frames. ], batch size: 28, lr: 4.58e-03, grad_scale: 8.0 +2023-03-21 08:10:34,862 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 08:10:36,423 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 1.808e+02 2.052e+02 2.359e+02 3.497e+02, threshold=4.104e+02, percent-clipped=0.0 +2023-03-21 08:10:40,639 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 +2023-03-21 08:10:45,296 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 08:10:59,254 INFO [train.py:901] (1/2) Epoch 36, batch 1750, loss[loss=0.1404, simple_loss=0.2237, pruned_loss=0.0286, over 7238.00 frames. ], tot_loss[loss=0.1326, simple_loss=0.2141, pruned_loss=0.02554, over 1443169.27 frames. ], batch size: 89, lr: 4.58e-03, grad_scale: 8.0 +2023-03-21 08:11:00,825 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100594.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:11:09,041 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 08:11:10,126 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 08:11:26,002 INFO [train.py:901] (1/2) Epoch 36, batch 1800, loss[loss=0.1417, simple_loss=0.2192, pruned_loss=0.03212, over 7254.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.2132, pruned_loss=0.0252, over 1441399.79 frames. ], batch size: 55, lr: 4.58e-03, grad_scale: 8.0 +2023-03-21 08:11:28,012 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.362e+02 1.832e+02 2.095e+02 2.409e+02 4.709e+02, threshold=4.189e+02, percent-clipped=2.0 +2023-03-21 08:11:29,378 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 +2023-03-21 08:11:32,099 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 08:11:37,354 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100663.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 08:11:43,869 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7410, 2.9858, 2.6755, 2.9518, 2.9471, 2.6444, 3.0022, 2.7716], + device='cuda:1'), covar=tensor([0.0824, 0.0507, 0.0884, 0.0879, 0.0706, 0.0685, 0.0943, 0.0789], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0057, 0.0066, 0.0058, 0.0055, 0.0060, 0.0055, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:11:45,255 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 08:11:51,208 INFO [train.py:901] (1/2) Epoch 36, batch 1850, loss[loss=0.1429, simple_loss=0.2272, pruned_loss=0.02924, over 7308.00 frames. ], tot_loss[loss=0.1322, simple_loss=0.2138, pruned_loss=0.02534, over 1443481.24 frames. ], batch size: 59, lr: 4.58e-03, grad_scale: 8.0 +2023-03-21 08:11:56,359 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 08:12:01,966 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100711.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 08:12:12,251 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100730.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 08:12:13,160 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 08:12:17,576 INFO [train.py:901] (1/2) Epoch 36, batch 1900, loss[loss=0.1317, simple_loss=0.2198, pruned_loss=0.02182, over 7293.00 frames. ], tot_loss[loss=0.1328, simple_loss=0.2141, pruned_loss=0.02569, over 1442166.44 frames. ], batch size: 86, lr: 4.58e-03, grad_scale: 8.0 +2023-03-21 08:12:19,529 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.072e+02 1.719e+02 1.937e+02 2.500e+02 4.543e+02, threshold=3.873e+02, percent-clipped=1.0 +2023-03-21 08:12:26,087 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100758.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 08:12:30,130 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9208, 3.6874, 3.5720, 3.6160, 3.5674, 3.5318, 3.7947, 3.4119], + device='cuda:1'), covar=tensor([0.0131, 0.0171, 0.0133, 0.0205, 0.0446, 0.0119, 0.0159, 0.0179], + device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0101, 0.0101, 0.0090, 0.0175, 0.0106, 0.0104, 0.0110], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:12:36,501 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 08:12:38,133 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4127, 3.3758, 3.1423, 3.5926, 3.1937, 2.8170, 3.5065, 3.1298], + device='cuda:1'), covar=tensor([0.1157, 0.1268, 0.1068, 0.0965, 0.2271, 0.0962, 0.0925, 0.1459], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0057, 0.0066, 0.0058, 0.0055, 0.0060, 0.0055, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:12:43,114 INFO [train.py:901] (1/2) Epoch 36, batch 1950, loss[loss=0.1404, simple_loss=0.2189, pruned_loss=0.03092, over 7285.00 frames. ], tot_loss[loss=0.1328, simple_loss=0.2139, pruned_loss=0.02589, over 1440974.74 frames. ], batch size: 47, lr: 4.57e-03, grad_scale: 8.0 +2023-03-21 08:12:43,278 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100791.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 08:12:47,626 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 08:12:53,129 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 08:12:54,070 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 08:13:00,221 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1648, 2.3269, 2.4396, 2.1050, 2.3888, 2.2574, 1.8973, 1.8684], + device='cuda:1'), covar=tensor([0.0472, 0.0500, 0.0268, 0.0243, 0.0282, 0.0431, 0.0440, 0.0332], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0038, 0.0039, 0.0038, 0.0035, 0.0036, 0.0041, 0.0040], + device='cuda:1'), out_proj_covar=tensor([9.8731e-05, 9.8065e-05, 9.7875e-05, 9.5130e-05, 9.3484e-05, 9.3996e-05, + 1.0216e-04, 1.0198e-04], device='cuda:1') +2023-03-21 08:13:09,060 INFO [train.py:901] (1/2) Epoch 36, batch 2000, loss[loss=0.1406, simple_loss=0.2238, pruned_loss=0.0287, over 7307.00 frames. ], tot_loss[loss=0.1328, simple_loss=0.214, pruned_loss=0.02581, over 1441953.64 frames. ], batch size: 59, lr: 4.57e-03, grad_scale: 8.0 +2023-03-21 08:13:10,525 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 08:13:10,966 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 1.685e+02 2.083e+02 2.423e+02 3.293e+02, threshold=4.166e+02, percent-clipped=0.0 +2023-03-21 08:13:20,596 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 08:13:28,716 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 08:13:35,221 INFO [train.py:901] (1/2) Epoch 36, batch 2050, loss[loss=0.1596, simple_loss=0.2301, pruned_loss=0.04453, over 7237.00 frames. ], tot_loss[loss=0.133, simple_loss=0.2144, pruned_loss=0.0258, over 1442782.22 frames. ], batch size: 55, lr: 4.57e-03, grad_scale: 8.0 +2023-03-21 08:13:36,803 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100894.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:13:39,395 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8795, 3.7146, 3.8525, 3.7616, 3.5938, 3.5240, 4.1106, 3.1083], + device='cuda:1'), covar=tensor([0.0484, 0.0771, 0.0505, 0.0673, 0.0793, 0.0947, 0.0635, 0.1816], + device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0338, 0.0274, 0.0354, 0.0293, 0.0290, 0.0346, 0.0252], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 08:13:58,099 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9142, 3.1114, 3.9259, 3.7977, 3.9532, 3.9544, 3.9682, 3.8984], + device='cuda:1'), covar=tensor([0.0032, 0.0127, 0.0031, 0.0033, 0.0028, 0.0027, 0.0036, 0.0044], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0067, 0.0056, 0.0055, 0.0053, 0.0059, 0.0047, 0.0074], + device='cuda:1'), out_proj_covar=tensor([8.1076e-05, 1.3838e-04, 1.0222e-04, 9.7009e-05, 9.3069e-05, 1.0467e-04, + 9.1057e-05, 1.3942e-04], device='cuda:1') +2023-03-21 08:14:00,394 INFO [train.py:901] (1/2) Epoch 36, batch 2100, loss[loss=0.1345, simple_loss=0.2173, pruned_loss=0.02588, over 7250.00 frames. ], tot_loss[loss=0.1326, simple_loss=0.2137, pruned_loss=0.02569, over 1443201.09 frames. ], batch size: 89, lr: 4.57e-03, grad_scale: 8.0 +2023-03-21 08:14:00,937 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=100942.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:14:01,393 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 08:14:02,344 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.257e+02 1.697e+02 1.994e+02 2.403e+02 4.710e+02, threshold=3.988e+02, percent-clipped=1.0 +2023-03-21 08:14:03,880 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 08:14:04,997 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8473, 3.1164, 3.9006, 3.8966, 3.8959, 3.9800, 3.9455, 3.9201], + device='cuda:1'), covar=tensor([0.0037, 0.0134, 0.0032, 0.0037, 0.0030, 0.0027, 0.0042, 0.0047], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0067, 0.0055, 0.0055, 0.0053, 0.0059, 0.0047, 0.0074], + device='cuda:1'), out_proj_covar=tensor([8.0924e-05, 1.3810e-04, 1.0221e-04, 9.6701e-05, 9.2900e-05, 1.0437e-04, + 9.1075e-05, 1.3924e-04], device='cuda:1') +2023-03-21 08:14:06,028 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1861, 2.3398, 2.5381, 2.0332, 2.2670, 2.4609, 2.0859, 2.0343], + device='cuda:1'), covar=tensor([0.0652, 0.0461, 0.0395, 0.0291, 0.0783, 0.0464, 0.0343, 0.0350], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0038, 0.0039, 0.0038, 0.0036, 0.0037, 0.0042, 0.0041], + device='cuda:1'), out_proj_covar=tensor([9.9122e-05, 9.9077e-05, 9.8919e-05, 9.6665e-05, 9.4510e-05, 9.4704e-05, + 1.0335e-04, 1.0285e-04], device='cuda:1') +2023-03-21 08:14:26,898 INFO [train.py:901] (1/2) Epoch 36, batch 2150, loss[loss=0.1205, simple_loss=0.2096, pruned_loss=0.01568, over 7287.00 frames. ], tot_loss[loss=0.132, simple_loss=0.2133, pruned_loss=0.02535, over 1441103.83 frames. ], batch size: 68, lr: 4.57e-03, grad_scale: 8.0 +2023-03-21 08:14:28,848 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 +2023-03-21 08:14:50,421 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 08:14:53,043 INFO [train.py:901] (1/2) Epoch 36, batch 2200, loss[loss=0.1161, simple_loss=0.1998, pruned_loss=0.01618, over 7327.00 frames. ], tot_loss[loss=0.1316, simple_loss=0.2132, pruned_loss=0.025, over 1441387.80 frames. ], batch size: 44, lr: 4.57e-03, grad_scale: 8.0 +2023-03-21 08:14:55,019 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.755e+02 1.986e+02 2.437e+02 3.637e+02, threshold=3.971e+02, percent-clipped=0.0 +2023-03-21 08:15:01,678 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101058.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:15:16,331 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101086.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 08:15:18,794 INFO [train.py:901] (1/2) Epoch 36, batch 2250, loss[loss=0.133, simple_loss=0.2158, pruned_loss=0.02511, over 7256.00 frames. ], tot_loss[loss=0.1316, simple_loss=0.213, pruned_loss=0.02508, over 1442983.65 frames. ], batch size: 77, lr: 4.57e-03, grad_scale: 4.0 +2023-03-21 08:15:26,348 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 08:15:26,359 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 08:15:26,398 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=101106.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:15:26,930 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101107.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:15:30,995 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8673, 3.1830, 2.7038, 3.0122, 3.0856, 2.7592, 3.2323, 2.8372], + device='cuda:1'), covar=tensor([0.0836, 0.0804, 0.0955, 0.1241, 0.0552, 0.0528, 0.0880, 0.1222], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0058, 0.0067, 0.0059, 0.0056, 0.0061, 0.0056, 0.0054], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:15:36,313 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 +2023-03-21 08:15:38,981 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 08:15:44,463 INFO [train.py:901] (1/2) Epoch 36, batch 2300, loss[loss=0.1437, simple_loss=0.2243, pruned_loss=0.03154, over 7214.00 frames. ], tot_loss[loss=0.1322, simple_loss=0.2132, pruned_loss=0.02559, over 1442392.98 frames. ], batch size: 93, lr: 4.57e-03, grad_scale: 4.0 +2023-03-21 08:15:45,538 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6694, 5.1516, 5.2045, 5.1188, 4.9851, 4.7783, 5.2342, 5.0844], + device='cuda:1'), covar=tensor([0.0427, 0.0356, 0.0328, 0.0480, 0.0283, 0.0319, 0.0266, 0.0375], + device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0250, 0.0194, 0.0195, 0.0154, 0.0227, 0.0204, 0.0147], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 08:15:46,985 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.343e+02 1.779e+02 2.010e+02 2.445e+02 5.360e+02, threshold=4.019e+02, percent-clipped=2.0 +2023-03-21 08:15:54,997 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 +2023-03-21 08:15:58,820 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101168.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:16:01,354 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8652, 3.7186, 2.9249, 3.4771, 2.6719, 2.1175, 1.7217, 3.7924], + device='cuda:1'), covar=tensor([0.0075, 0.0082, 0.0222, 0.0099, 0.0273, 0.0770, 0.0770, 0.0075], + device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0091, 0.0113, 0.0094, 0.0128, 0.0135, 0.0130, 0.0102], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 08:16:10,345 INFO [train.py:901] (1/2) Epoch 36, batch 2350, loss[loss=0.1402, simple_loss=0.232, pruned_loss=0.02414, over 7232.00 frames. ], tot_loss[loss=0.132, simple_loss=0.2132, pruned_loss=0.02541, over 1441975.36 frames. ], batch size: 93, lr: 4.57e-03, grad_scale: 4.0 +2023-03-21 08:16:25,869 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 08:16:32,817 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 08:16:36,971 INFO [train.py:901] (1/2) Epoch 36, batch 2400, loss[loss=0.1256, simple_loss=0.2159, pruned_loss=0.01769, over 7277.00 frames. ], tot_loss[loss=0.1322, simple_loss=0.2133, pruned_loss=0.02558, over 1442593.23 frames. ], batch size: 70, lr: 4.56e-03, grad_scale: 8.0 +2023-03-21 08:16:39,422 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.253e+02 1.751e+02 2.088e+02 2.465e+02 5.704e+02, threshold=4.177e+02, percent-clipped=6.0 +2023-03-21 08:16:44,540 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 08:16:47,519 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 08:17:02,200 INFO [train.py:901] (1/2) Epoch 36, batch 2450, loss[loss=0.1235, simple_loss=0.2083, pruned_loss=0.01938, over 7300.00 frames. ], tot_loss[loss=0.132, simple_loss=0.213, pruned_loss=0.02555, over 1439794.69 frames. ], batch size: 86, lr: 4.56e-03, grad_scale: 4.0 +2023-03-21 08:17:03,961 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-03-21 08:17:13,767 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 08:17:28,468 INFO [train.py:901] (1/2) Epoch 36, batch 2500, loss[loss=0.1352, simple_loss=0.2142, pruned_loss=0.0281, over 7329.00 frames. ], tot_loss[loss=0.1324, simple_loss=0.2133, pruned_loss=0.02571, over 1440515.63 frames. ], batch size: 83, lr: 4.56e-03, grad_scale: 4.0 +2023-03-21 08:17:31,516 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.249e+02 1.796e+02 2.233e+02 2.673e+02 5.552e+02, threshold=4.467e+02, percent-clipped=2.0 +2023-03-21 08:17:39,176 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 08:17:51,291 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7704, 2.3182, 2.9180, 2.7537, 2.8246, 2.6796, 2.3018, 2.9150], + device='cuda:1'), covar=tensor([0.1787, 0.0988, 0.1420, 0.1343, 0.0917, 0.1154, 0.2641, 0.1258], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0068, 0.0051, 0.0050, 0.0050, 0.0049, 0.0068, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 08:17:51,781 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101386.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 08:17:54,140 INFO [train.py:901] (1/2) Epoch 36, batch 2550, loss[loss=0.1086, simple_loss=0.19, pruned_loss=0.01361, over 7167.00 frames. ], tot_loss[loss=0.1323, simple_loss=0.2135, pruned_loss=0.02562, over 1443977.76 frames. ], batch size: 41, lr: 4.56e-03, grad_scale: 4.0 +2023-03-21 08:18:10,209 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0623, 4.6307, 4.3509, 5.0320, 4.8261, 5.0107, 4.3840, 4.5665], + device='cuda:1'), covar=tensor([0.0841, 0.2448, 0.2447, 0.1151, 0.1056, 0.1174, 0.0739, 0.1145], + device='cuda:1'), in_proj_covar=tensor([0.0259, 0.0388, 0.0296, 0.0305, 0.0229, 0.0364, 0.0224, 0.0271], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:18:16,603 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=101434.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 08:18:19,946 INFO [train.py:901] (1/2) Epoch 36, batch 2600, loss[loss=0.1395, simple_loss=0.2212, pruned_loss=0.02892, over 7350.00 frames. ], tot_loss[loss=0.1321, simple_loss=0.2136, pruned_loss=0.02531, over 1444465.85 frames. ], batch size: 73, lr: 4.56e-03, grad_scale: 4.0 +2023-03-21 08:18:22,803 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.775e+02 2.071e+02 2.453e+02 3.538e+02, threshold=4.142e+02, percent-clipped=0.0 +2023-03-21 08:18:30,709 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101463.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:18:44,269 INFO [train.py:901] (1/2) Epoch 36, batch 2650, loss[loss=0.1426, simple_loss=0.2227, pruned_loss=0.03131, over 7270.00 frames. ], tot_loss[loss=0.1321, simple_loss=0.2132, pruned_loss=0.02551, over 1443429.18 frames. ], batch size: 47, lr: 4.56e-03, grad_scale: 4.0 +2023-03-21 08:18:52,252 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1370, 4.6080, 4.6597, 4.5909, 4.5758, 4.1590, 4.6628, 4.5279], + device='cuda:1'), covar=tensor([0.0458, 0.0379, 0.0354, 0.0484, 0.0305, 0.0444, 0.0325, 0.0437], + device='cuda:1'), in_proj_covar=tensor([0.0214, 0.0251, 0.0195, 0.0198, 0.0154, 0.0228, 0.0203, 0.0149], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 08:19:08,895 INFO [train.py:901] (1/2) Epoch 36, batch 2700, loss[loss=0.1518, simple_loss=0.2329, pruned_loss=0.03537, over 7241.00 frames. ], tot_loss[loss=0.1316, simple_loss=0.2128, pruned_loss=0.0252, over 1442880.76 frames. ], batch size: 55, lr: 4.56e-03, grad_scale: 4.0 +2023-03-21 08:19:11,761 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.222e+02 1.662e+02 1.940e+02 2.331e+02 3.530e+02, threshold=3.880e+02, percent-clipped=0.0 +2023-03-21 08:19:13,613 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-03-21 08:19:20,508 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 +2023-03-21 08:19:20,773 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2355, 4.4023, 4.1414, 4.4528, 4.0183, 4.4083, 4.7043, 4.6904], + device='cuda:1'), covar=tensor([0.0183, 0.0130, 0.0220, 0.0133, 0.0306, 0.0322, 0.0242, 0.0186], + device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0126, 0.0118, 0.0124, 0.0113, 0.0104, 0.0099, 0.0101], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:19:34,101 INFO [train.py:901] (1/2) Epoch 36, batch 2750, loss[loss=0.1276, simple_loss=0.2077, pruned_loss=0.02371, over 7354.00 frames. ], tot_loss[loss=0.1317, simple_loss=0.2128, pruned_loss=0.02532, over 1444022.06 frames. ], batch size: 63, lr: 4.56e-03, grad_scale: 4.0 +2023-03-21 08:19:58,925 INFO [train.py:901] (1/2) Epoch 36, batch 2800, loss[loss=0.1279, simple_loss=0.2066, pruned_loss=0.02463, over 7349.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.2128, pruned_loss=0.02541, over 1446766.16 frames. ], batch size: 51, lr: 4.56e-03, grad_scale: 8.0 +2023-03-21 08:20:01,833 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.705e+02 1.985e+02 2.296e+02 4.387e+02, threshold=3.971e+02, percent-clipped=1.0 +2023-03-21 08:20:24,101 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 08:20:24,165 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 08:20:24,190 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 08:20:24,206 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 08:20:24,379 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 08:20:24,395 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 08:20:24,406 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 08:20:24,507 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 08:20:24,581 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 08:20:24,595 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 08:20:24,612 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 08:20:25,118 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 08:20:25,124 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 08:20:25,131 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 08:20:25,273 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 08:20:25,275 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 08:20:25,333 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 08:20:30,579 INFO [train.py:901] (1/2) Epoch 37, batch 0, loss[loss=0.1411, simple_loss=0.2268, pruned_loss=0.02767, over 7132.00 frames. ], tot_loss[loss=0.1411, simple_loss=0.2268, pruned_loss=0.02767, over 7132.00 frames. ], batch size: 98, lr: 4.49e-03, grad_scale: 8.0 +2023-03-21 08:20:30,579 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 08:20:44,274 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.9806, 5.2674, 5.3314, 5.2587, 4.8735, 4.8806, 5.3244, 4.9959], + device='cuda:1'), covar=tensor([0.0429, 0.0334, 0.0326, 0.0449, 0.0362, 0.0344, 0.0287, 0.0507], + device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0251, 0.0194, 0.0196, 0.0153, 0.0227, 0.0202, 0.0148], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 08:20:47,002 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6606, 4.2721, 4.0045, 4.7317, 4.4661, 4.7581, 4.4376, 4.5299], + device='cuda:1'), covar=tensor([0.0755, 0.2135, 0.1856, 0.1019, 0.0867, 0.0875, 0.0523, 0.0857], + device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0383, 0.0291, 0.0300, 0.0226, 0.0357, 0.0221, 0.0266], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:20:57,445 INFO [train.py:935] (1/2) Epoch 37, validation: loss=0.1643, simple_loss=0.2561, pruned_loss=0.03627, over 1622729.00 frames. +2023-03-21 08:20:57,445 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 08:21:04,433 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 08:21:07,073 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101684.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:21:15,524 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 08:21:22,136 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 08:21:22,595 INFO [train.py:901] (1/2) Epoch 37, batch 50, loss[loss=0.1602, simple_loss=0.246, pruned_loss=0.03717, over 6763.00 frames. ], tot_loss[loss=0.1291, simple_loss=0.2107, pruned_loss=0.02379, over 325206.46 frames. ], batch size: 107, lr: 4.49e-03, grad_scale: 8.0 +2023-03-21 08:21:24,133 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 08:21:25,966 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 +2023-03-21 08:21:27,121 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 08:21:37,746 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101745.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:21:38,565 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.229e+02 1.749e+02 1.978e+02 2.334e+02 3.744e+02, threshold=3.956e+02, percent-clipped=0.0 +2023-03-21 08:21:43,885 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 08:21:44,406 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 08:21:47,968 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101763.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:21:48,814 INFO [train.py:901] (1/2) Epoch 37, batch 100, loss[loss=0.1328, simple_loss=0.2151, pruned_loss=0.02524, over 7281.00 frames. ], tot_loss[loss=0.1309, simple_loss=0.2125, pruned_loss=0.02468, over 573767.10 frames. ], batch size: 66, lr: 4.49e-03, grad_scale: 8.0 +2023-03-21 08:22:12,330 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=101811.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:22:14,205 INFO [train.py:901] (1/2) Epoch 37, batch 150, loss[loss=0.1692, simple_loss=0.2368, pruned_loss=0.05084, over 7247.00 frames. ], tot_loss[loss=0.1325, simple_loss=0.2136, pruned_loss=0.02572, over 762633.22 frames. ], batch size: 55, lr: 4.49e-03, grad_scale: 8.0 +2023-03-21 08:22:19,915 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7477, 3.9001, 3.7498, 3.9337, 3.5491, 3.8567, 4.1848, 4.1609], + device='cuda:1'), covar=tensor([0.0219, 0.0153, 0.0191, 0.0161, 0.0375, 0.0345, 0.0200, 0.0184], + device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0125, 0.0117, 0.0123, 0.0112, 0.0103, 0.0097, 0.0100], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:22:31,469 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.264e+02 1.790e+02 2.048e+02 2.419e+02 6.579e+02, threshold=4.096e+02, percent-clipped=3.0 +2023-03-21 08:22:40,653 INFO [train.py:901] (1/2) Epoch 37, batch 200, loss[loss=0.1402, simple_loss=0.2258, pruned_loss=0.02726, over 7270.00 frames. ], tot_loss[loss=0.1321, simple_loss=0.213, pruned_loss=0.02562, over 912233.85 frames. ], batch size: 52, lr: 4.49e-03, grad_scale: 8.0 +2023-03-21 08:22:46,226 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 08:22:50,711 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 08:22:51,394 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7807, 1.4714, 2.0872, 2.3028, 2.1433, 2.0610, 1.8784, 2.2678], + device='cuda:1'), covar=tensor([0.3274, 0.3890, 0.1098, 0.0728, 0.1601, 0.1425, 0.1496, 0.1528], + device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0076, 0.0068, 0.0062, 0.0060, 0.0061, 0.0101, 0.0064], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 08:22:56,870 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 08:23:01,737 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 +2023-03-21 08:23:05,974 INFO [train.py:901] (1/2) Epoch 37, batch 250, loss[loss=0.1203, simple_loss=0.2015, pruned_loss=0.01954, over 7282.00 frames. ], tot_loss[loss=0.1315, simple_loss=0.2127, pruned_loss=0.02518, over 1031275.37 frames. ], batch size: 68, lr: 4.49e-03, grad_scale: 8.0 +2023-03-21 08:23:08,543 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 08:23:08,644 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101920.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 08:23:23,076 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.200e+02 1.857e+02 2.138e+02 2.562e+02 4.805e+02, threshold=4.275e+02, percent-clipped=2.0 +2023-03-21 08:23:31,149 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 08:23:32,135 INFO [train.py:901] (1/2) Epoch 37, batch 300, loss[loss=0.1461, simple_loss=0.2309, pruned_loss=0.03061, over 7286.00 frames. ], tot_loss[loss=0.1312, simple_loss=0.2126, pruned_loss=0.02491, over 1123287.63 frames. ], batch size: 66, lr: 4.49e-03, grad_scale: 8.0 +2023-03-21 08:23:40,181 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 08:23:40,328 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101981.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 08:23:43,312 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101987.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:23:57,670 INFO [train.py:901] (1/2) Epoch 37, batch 350, loss[loss=0.1397, simple_loss=0.2232, pruned_loss=0.02813, over 7237.00 frames. ], tot_loss[loss=0.1314, simple_loss=0.2125, pruned_loss=0.02517, over 1192556.87 frames. ], batch size: 93, lr: 4.49e-03, grad_scale: 8.0 +2023-03-21 08:24:11,548 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102040.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:24:14,987 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.770e+02 2.044e+02 2.367e+02 3.943e+02, threshold=4.088e+02, percent-clipped=0.0 +2023-03-21 08:24:15,502 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 08:24:15,617 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102048.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:24:17,664 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2903, 3.2123, 2.4804, 3.7844, 2.8097, 3.3092, 1.7976, 2.7074], + device='cuda:1'), covar=tensor([0.0390, 0.0712, 0.2318, 0.0427, 0.0454, 0.0551, 0.3288, 0.1692], + device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0260, 0.0284, 0.0268, 0.0272, 0.0266, 0.0237, 0.0262], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:24:19,436 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 +2023-03-21 08:24:23,971 INFO [train.py:901] (1/2) Epoch 37, batch 400, loss[loss=0.1228, simple_loss=0.2017, pruned_loss=0.02191, over 7274.00 frames. ], tot_loss[loss=0.1317, simple_loss=0.2132, pruned_loss=0.02517, over 1250479.21 frames. ], batch size: 64, lr: 4.48e-03, grad_scale: 8.0 +2023-03-21 08:24:29,115 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9401, 3.3265, 3.8748, 4.0953, 4.0442, 3.9411, 4.1196, 3.8965], + device='cuda:1'), covar=tensor([0.0033, 0.0113, 0.0033, 0.0028, 0.0027, 0.0035, 0.0031, 0.0047], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0068, 0.0056, 0.0055, 0.0054, 0.0060, 0.0047, 0.0075], + device='cuda:1'), out_proj_covar=tensor([8.1994e-05, 1.3966e-04, 1.0394e-04, 9.7020e-05, 9.4078e-05, 1.0548e-04, + 9.0872e-05, 1.4041e-04], device='cuda:1') +2023-03-21 08:24:41,070 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.69 vs. limit=5.0 +2023-03-21 08:24:50,339 INFO [train.py:901] (1/2) Epoch 37, batch 450, loss[loss=0.1106, simple_loss=0.1971, pruned_loss=0.01211, over 7157.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.2131, pruned_loss=0.02525, over 1293148.59 frames. ], batch size: 41, lr: 4.48e-03, grad_scale: 8.0 +2023-03-21 08:24:54,638 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 +2023-03-21 08:24:56,921 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 08:24:57,413 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 08:25:06,272 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.299e+02 1.769e+02 2.012e+02 2.507e+02 5.212e+02, threshold=4.024e+02, percent-clipped=1.0 +2023-03-21 08:25:07,415 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1053, 3.7441, 3.7375, 3.8447, 3.7846, 3.6794, 3.9354, 3.4532], + device='cuda:1'), covar=tensor([0.0154, 0.0189, 0.0130, 0.0173, 0.0395, 0.0111, 0.0168, 0.0195], + device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0102, 0.0102, 0.0089, 0.0177, 0.0107, 0.0106, 0.0110], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:25:15,492 INFO [train.py:901] (1/2) Epoch 37, batch 500, loss[loss=0.1401, simple_loss=0.2221, pruned_loss=0.02904, over 7346.00 frames. ], tot_loss[loss=0.132, simple_loss=0.2135, pruned_loss=0.02527, over 1327868.17 frames. ], batch size: 54, lr: 4.48e-03, grad_scale: 8.0 +2023-03-21 08:25:17,824 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 +2023-03-21 08:25:30,221 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 08:25:31,741 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 08:25:32,215 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 08:25:35,109 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 08:25:39,663 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 08:25:42,198 INFO [train.py:901] (1/2) Epoch 37, batch 550, loss[loss=0.1301, simple_loss=0.216, pruned_loss=0.02215, over 7244.00 frames. ], tot_loss[loss=0.1315, simple_loss=0.2128, pruned_loss=0.02512, over 1354512.01 frames. ], batch size: 45, lr: 4.48e-03, grad_scale: 8.0 +2023-03-21 08:25:50,329 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1669, 4.3404, 4.0985, 4.3429, 3.9296, 4.3440, 4.6373, 4.6043], + device='cuda:1'), covar=tensor([0.0224, 0.0157, 0.0232, 0.0172, 0.0429, 0.0231, 0.0239, 0.0215], + device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0126, 0.0118, 0.0124, 0.0112, 0.0103, 0.0098, 0.0101], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:25:50,738 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 08:25:58,230 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.288e+02 1.789e+02 2.159e+02 2.504e+02 4.340e+02, threshold=4.318e+02, percent-clipped=1.0 +2023-03-21 08:25:58,781 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 08:26:01,828 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 08:26:07,368 INFO [train.py:901] (1/2) Epoch 37, batch 600, loss[loss=0.1351, simple_loss=0.2197, pruned_loss=0.02519, over 7279.00 frames. ], tot_loss[loss=0.1317, simple_loss=0.2132, pruned_loss=0.02506, over 1374832.18 frames. ], batch size: 66, lr: 4.48e-03, grad_scale: 8.0 +2023-03-21 08:26:08,397 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 08:26:13,682 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102276.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 08:26:25,447 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 08:26:33,500 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 08:26:33,990 INFO [train.py:901] (1/2) Epoch 37, batch 650, loss[loss=0.1424, simple_loss=0.223, pruned_loss=0.03085, over 7243.00 frames. ], tot_loss[loss=0.1321, simple_loss=0.2139, pruned_loss=0.02522, over 1390698.39 frames. ], batch size: 89, lr: 4.48e-03, grad_scale: 8.0 +2023-03-21 08:26:36,951 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-21 08:26:46,866 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102340.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:26:48,408 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102343.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:26:49,879 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 08:26:50,341 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.201e+02 1.711e+02 1.994e+02 2.328e+02 4.780e+02, threshold=3.988e+02, percent-clipped=1.0 +2023-03-21 08:26:59,310 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 08:27:00,873 INFO [train.py:901] (1/2) Epoch 37, batch 700, loss[loss=0.1094, simple_loss=0.1716, pruned_loss=0.02359, over 5824.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.2132, pruned_loss=0.02519, over 1401346.79 frames. ], batch size: 25, lr: 4.48e-03, grad_scale: 8.0 +2023-03-21 08:27:12,410 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-21 08:27:12,698 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=102388.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:27:19,395 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6478, 2.8898, 2.5556, 2.7491, 2.7267, 2.4731, 2.8702, 2.6920], + device='cuda:1'), covar=tensor([0.0737, 0.0848, 0.0732, 0.1002, 0.1012, 0.0704, 0.0555, 0.0858], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0056, 0.0064, 0.0057, 0.0054, 0.0059, 0.0055, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:27:23,284 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 08:27:24,267 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 08:27:26,745 INFO [train.py:901] (1/2) Epoch 37, batch 750, loss[loss=0.1358, simple_loss=0.2197, pruned_loss=0.02595, over 7293.00 frames. ], tot_loss[loss=0.132, simple_loss=0.2134, pruned_loss=0.02534, over 1407735.22 frames. ], batch size: 68, lr: 4.48e-03, grad_scale: 8.0 +2023-03-21 08:27:32,511 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8460, 2.8202, 2.3289, 3.5231, 2.4364, 3.0896, 1.6000, 2.3082], + device='cuda:1'), covar=tensor([0.0487, 0.0677, 0.2747, 0.0703, 0.0443, 0.0600, 0.3895, 0.1854], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0259, 0.0282, 0.0268, 0.0272, 0.0267, 0.0236, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:27:37,937 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 08:27:42,817 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 08:27:43,787 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.733e+02 2.069e+02 2.404e+02 6.127e+02, threshold=4.137e+02, percent-clipped=2.0 +2023-03-21 08:27:49,063 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 08:27:50,551 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 08:27:53,668 INFO [train.py:901] (1/2) Epoch 37, batch 800, loss[loss=0.1236, simple_loss=0.1987, pruned_loss=0.02422, over 7318.00 frames. ], tot_loss[loss=0.1319, simple_loss=0.2129, pruned_loss=0.02541, over 1411995.05 frames. ], batch size: 49, lr: 4.48e-03, grad_scale: 8.0 +2023-03-21 08:28:00,675 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 08:28:18,808 INFO [train.py:901] (1/2) Epoch 37, batch 850, loss[loss=0.1229, simple_loss=0.213, pruned_loss=0.0164, over 7293.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.213, pruned_loss=0.02534, over 1421247.19 frames. ], batch size: 80, lr: 4.47e-03, grad_scale: 8.0 +2023-03-21 08:28:19,333 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 08:28:19,761 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 08:28:21,461 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1792, 3.0371, 2.3611, 3.6265, 2.6735, 3.1869, 1.6614, 2.5827], + device='cuda:1'), covar=tensor([0.0460, 0.0820, 0.2500, 0.0612, 0.0483, 0.0635, 0.3512, 0.1600], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0259, 0.0282, 0.0268, 0.0271, 0.0266, 0.0236, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:28:25,897 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 08:28:29,394 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 08:28:35,939 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.128e+02 1.881e+02 2.184e+02 2.686e+02 4.257e+02, threshold=4.368e+02, percent-clipped=1.0 +2023-03-21 08:28:45,135 INFO [train.py:901] (1/2) Epoch 37, batch 900, loss[loss=0.1417, simple_loss=0.2243, pruned_loss=0.0296, over 7310.00 frames. ], tot_loss[loss=0.1321, simple_loss=0.2132, pruned_loss=0.02553, over 1427753.99 frames. ], batch size: 83, lr: 4.47e-03, grad_scale: 8.0 +2023-03-21 08:28:50,875 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102576.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 08:28:54,544 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 +2023-03-21 08:29:07,814 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 08:29:11,424 INFO [train.py:901] (1/2) Epoch 37, batch 950, loss[loss=0.1433, simple_loss=0.2275, pruned_loss=0.02956, over 7321.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.213, pruned_loss=0.02534, over 1429371.37 frames. ], batch size: 83, lr: 4.47e-03, grad_scale: 8.0 +2023-03-21 08:29:14,715 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6228, 4.1590, 4.1700, 4.3151, 4.2718, 4.1598, 4.5067, 3.9907], + device='cuda:1'), covar=tensor([0.0140, 0.0174, 0.0114, 0.0151, 0.0401, 0.0104, 0.0128, 0.0164], + device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0100, 0.0101, 0.0089, 0.0175, 0.0106, 0.0103, 0.0110], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:29:16,736 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=102624.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 08:29:26,317 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102643.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:29:28,162 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.204e+02 1.662e+02 1.972e+02 2.353e+02 5.734e+02, threshold=3.944e+02, percent-clipped=1.0 +2023-03-21 08:29:31,741 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 08:29:37,267 INFO [train.py:901] (1/2) Epoch 37, batch 1000, loss[loss=0.1519, simple_loss=0.2436, pruned_loss=0.03012, over 6788.00 frames. ], tot_loss[loss=0.1315, simple_loss=0.2125, pruned_loss=0.02522, over 1429141.51 frames. ], batch size: 107, lr: 4.47e-03, grad_scale: 8.0 +2023-03-21 08:29:50,344 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=102691.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:29:52,276 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 08:30:03,335 INFO [train.py:901] (1/2) Epoch 37, batch 1050, loss[loss=0.1232, simple_loss=0.2075, pruned_loss=0.01946, over 7272.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.213, pruned_loss=0.02533, over 1434157.18 frames. ], batch size: 66, lr: 4.47e-03, grad_scale: 8.0 +2023-03-21 08:30:14,450 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 08:30:18,484 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 08:30:19,451 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.402e+02 1.793e+02 2.007e+02 2.334e+02 3.523e+02, threshold=4.014e+02, percent-clipped=0.0 +2023-03-21 08:30:28,471 INFO [train.py:901] (1/2) Epoch 37, batch 1100, loss[loss=0.1425, simple_loss=0.2211, pruned_loss=0.03192, over 7286.00 frames. ], tot_loss[loss=0.1319, simple_loss=0.213, pruned_loss=0.02542, over 1434871.60 frames. ], batch size: 57, lr: 4.47e-03, grad_scale: 8.0 +2023-03-21 08:30:44,958 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 +2023-03-21 08:30:49,089 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 08:30:49,101 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 08:30:55,023 INFO [train.py:901] (1/2) Epoch 37, batch 1150, loss[loss=0.1341, simple_loss=0.2199, pruned_loss=0.02418, over 7285.00 frames. ], tot_loss[loss=0.1307, simple_loss=0.2118, pruned_loss=0.02484, over 1434844.62 frames. ], batch size: 70, lr: 4.47e-03, grad_scale: 8.0 +2023-03-21 08:31:02,021 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 08:31:02,491 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 08:31:10,925 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.364e+02 1.647e+02 1.990e+02 2.353e+02 6.242e+02, threshold=3.979e+02, percent-clipped=4.0 +2023-03-21 08:31:13,095 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102851.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:31:19,888 INFO [train.py:901] (1/2) Epoch 37, batch 1200, loss[loss=0.1298, simple_loss=0.2021, pruned_loss=0.02878, over 7230.00 frames. ], tot_loss[loss=0.131, simple_loss=0.2122, pruned_loss=0.02491, over 1437450.85 frames. ], batch size: 45, lr: 4.47e-03, grad_scale: 8.0 +2023-03-21 08:31:31,520 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102887.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:31:34,040 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 08:31:40,264 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6124, 2.8422, 2.5200, 2.8510, 2.7292, 2.4578, 2.7682, 2.6690], + device='cuda:1'), covar=tensor([0.0698, 0.1066, 0.1058, 0.0996, 0.0956, 0.0780, 0.0784, 0.0875], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0057, 0.0065, 0.0057, 0.0055, 0.0060, 0.0055, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:31:42,263 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5791, 4.1815, 3.9923, 4.5984, 4.3658, 4.5160, 3.8943, 4.2202], + device='cuda:1'), covar=tensor([0.0847, 0.2474, 0.2226, 0.1084, 0.0955, 0.1258, 0.0925, 0.1169], + device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0384, 0.0294, 0.0303, 0.0227, 0.0365, 0.0223, 0.0270], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:31:42,521 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-03-21 08:31:44,827 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102912.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:31:46,203 INFO [train.py:901] (1/2) Epoch 37, batch 1250, loss[loss=0.1129, simple_loss=0.193, pruned_loss=0.01639, over 7181.00 frames. ], tot_loss[loss=0.1305, simple_loss=0.2116, pruned_loss=0.02472, over 1437792.55 frames. ], batch size: 39, lr: 4.47e-03, grad_scale: 8.0 +2023-03-21 08:31:57,761 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 08:32:02,203 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.332e+02 1.760e+02 2.143e+02 2.522e+02 3.461e+02, threshold=4.287e+02, percent-clipped=0.0 +2023-03-21 08:32:02,738 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 08:32:02,874 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102948.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:32:04,256 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 08:32:07,704 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 +2023-03-21 08:32:08,962 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4285, 3.0313, 2.0457, 3.1383, 3.4563, 3.2556, 3.0744, 3.1042], + device='cuda:1'), covar=tensor([0.2069, 0.0898, 0.3976, 0.0595, 0.0318, 0.0249, 0.0450, 0.0369], + device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0228, 0.0245, 0.0254, 0.0195, 0.0194, 0.0213, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 08:32:11,326 INFO [train.py:901] (1/2) Epoch 37, batch 1300, loss[loss=0.1316, simple_loss=0.2151, pruned_loss=0.02408, over 7256.00 frames. ], tot_loss[loss=0.1306, simple_loss=0.2116, pruned_loss=0.02481, over 1438285.10 frames. ], batch size: 89, lr: 4.46e-03, grad_scale: 8.0 +2023-03-21 08:32:27,586 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 08:32:30,855 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 08:32:32,130 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 +2023-03-21 08:32:34,331 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 08:32:37,943 INFO [train.py:901] (1/2) Epoch 37, batch 1350, loss[loss=0.145, simple_loss=0.228, pruned_loss=0.03094, over 7220.00 frames. ], tot_loss[loss=0.1309, simple_loss=0.2119, pruned_loss=0.02496, over 1440034.35 frames. ], batch size: 93, lr: 4.46e-03, grad_scale: 8.0 +2023-03-21 08:32:44,120 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 08:32:44,876 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-03-21 08:32:53,470 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 +2023-03-21 08:32:54,029 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.199e+02 1.761e+02 2.055e+02 2.469e+02 4.318e+02, threshold=4.111e+02, percent-clipped=1.0 +2023-03-21 08:33:04,299 INFO [train.py:901] (1/2) Epoch 37, batch 1400, loss[loss=0.1309, simple_loss=0.2106, pruned_loss=0.02566, over 7248.00 frames. ], tot_loss[loss=0.1311, simple_loss=0.2121, pruned_loss=0.02503, over 1438935.97 frames. ], batch size: 47, lr: 4.46e-03, grad_scale: 8.0 +2023-03-21 08:33:16,399 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 08:33:21,118 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103098.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:33:29,590 INFO [train.py:901] (1/2) Epoch 37, batch 1450, loss[loss=0.1261, simple_loss=0.2079, pruned_loss=0.02219, over 7350.00 frames. ], tot_loss[loss=0.1314, simple_loss=0.2124, pruned_loss=0.02521, over 1440201.79 frames. ], batch size: 73, lr: 4.46e-03, grad_scale: 8.0 +2023-03-21 08:33:38,189 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103132.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:33:39,590 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 08:33:41,229 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-03-21 08:33:46,848 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.242e+02 1.709e+02 2.082e+02 2.511e+02 4.451e+02, threshold=4.164e+02, percent-clipped=1.0 +2023-03-21 08:33:53,038 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103159.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:33:55,874 INFO [train.py:901] (1/2) Epoch 37, batch 1500, loss[loss=0.1277, simple_loss=0.2118, pruned_loss=0.02176, over 7285.00 frames. ], tot_loss[loss=0.1322, simple_loss=0.213, pruned_loss=0.02569, over 1439338.24 frames. ], batch size: 70, lr: 4.46e-03, grad_scale: 8.0 +2023-03-21 08:33:56,405 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 08:34:10,045 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103193.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:34:17,274 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103207.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:34:17,876 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8203, 3.5403, 3.5701, 3.7591, 3.2761, 3.2035, 3.9380, 2.8520], + device='cuda:1'), covar=tensor([0.0376, 0.0457, 0.0624, 0.0634, 0.0814, 0.0954, 0.0715, 0.2134], + device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0334, 0.0269, 0.0350, 0.0288, 0.0284, 0.0340, 0.0248], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 08:34:19,711 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 08:34:21,184 INFO [train.py:901] (1/2) Epoch 37, batch 1550, loss[loss=0.1444, simple_loss=0.2358, pruned_loss=0.02653, over 7258.00 frames. ], tot_loss[loss=0.1319, simple_loss=0.2128, pruned_loss=0.02549, over 1439140.14 frames. ], batch size: 64, lr: 4.46e-03, grad_scale: 8.0 +2023-03-21 08:34:35,279 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3474, 3.5172, 2.5410, 3.8406, 3.1807, 3.4819, 1.7432, 2.5393], + device='cuda:1'), covar=tensor([0.0511, 0.0942, 0.2377, 0.0623, 0.0557, 0.0880, 0.3688, 0.1948], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0257, 0.0281, 0.0268, 0.0272, 0.0265, 0.0235, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:34:36,659 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103243.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:34:37,985 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 +2023-03-21 08:34:38,623 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.258e+02 1.785e+02 2.115e+02 2.545e+02 5.224e+02, threshold=4.230e+02, percent-clipped=3.0 +2023-03-21 08:34:47,674 INFO [train.py:901] (1/2) Epoch 37, batch 1600, loss[loss=0.1056, simple_loss=0.1676, pruned_loss=0.02182, over 5923.00 frames. ], tot_loss[loss=0.132, simple_loss=0.2128, pruned_loss=0.02561, over 1439666.25 frames. ], batch size: 25, lr: 4.46e-03, grad_scale: 8.0 +2023-03-21 08:34:51,703 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 08:34:52,692 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 08:34:53,315 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0612, 2.4047, 1.7441, 3.0256, 2.8245, 2.8838, 2.6679, 2.6996], + device='cuda:1'), covar=tensor([0.2367, 0.1207, 0.4367, 0.0796, 0.0375, 0.0333, 0.0417, 0.0398], + device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0230, 0.0249, 0.0257, 0.0196, 0.0195, 0.0215, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 08:34:55,116 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 08:35:05,234 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 08:35:10,136 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 08:35:13,218 INFO [train.py:901] (1/2) Epoch 37, batch 1650, loss[loss=0.1223, simple_loss=0.1898, pruned_loss=0.0274, over 6953.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.2126, pruned_loss=0.02549, over 1441078.36 frames. ], batch size: 35, lr: 4.46e-03, grad_scale: 16.0 +2023-03-21 08:35:19,376 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 08:35:30,007 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 1.789e+02 2.098e+02 2.571e+02 4.173e+02, threshold=4.196e+02, percent-clipped=0.0 +2023-03-21 08:35:36,028 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 08:35:39,053 INFO [train.py:901] (1/2) Epoch 37, batch 1700, loss[loss=0.1502, simple_loss=0.2287, pruned_loss=0.03582, over 7163.00 frames. ], tot_loss[loss=0.132, simple_loss=0.2128, pruned_loss=0.02563, over 1439273.62 frames. ], batch size: 98, lr: 4.46e-03, grad_scale: 16.0 +2023-03-21 08:35:39,592 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 08:35:39,681 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103366.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 08:35:50,575 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 08:35:52,223 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8624, 2.2652, 2.9056, 2.8891, 3.0061, 2.8065, 2.4076, 2.9210], + device='cuda:1'), covar=tensor([0.1373, 0.1066, 0.1077, 0.1240, 0.0793, 0.1063, 0.2184, 0.1296], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0067, 0.0050, 0.0049, 0.0050, 0.0048, 0.0068, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 08:36:04,144 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-03-21 08:36:05,739 INFO [train.py:901] (1/2) Epoch 37, batch 1750, loss[loss=0.1273, simple_loss=0.2089, pruned_loss=0.02283, over 7280.00 frames. ], tot_loss[loss=0.1319, simple_loss=0.2126, pruned_loss=0.02561, over 1441243.66 frames. ], batch size: 66, lr: 4.46e-03, grad_scale: 8.0 +2023-03-21 08:36:11,971 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103427.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 08:36:15,930 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 08:36:16,962 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 08:36:22,527 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.197e+02 1.731e+02 1.992e+02 2.394e+02 4.370e+02, threshold=3.985e+02, percent-clipped=1.0 +2023-03-21 08:36:25,696 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103454.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:36:31,133 INFO [train.py:901] (1/2) Epoch 37, batch 1800, loss[loss=0.1255, simple_loss=0.1964, pruned_loss=0.02728, over 7233.00 frames. ], tot_loss[loss=0.1319, simple_loss=0.2123, pruned_loss=0.02579, over 1440652.86 frames. ], batch size: 45, lr: 4.45e-03, grad_scale: 8.0 +2023-03-21 08:36:38,090 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 08:36:43,254 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103488.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:36:48,676 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3170, 2.3436, 2.4458, 3.4128, 1.8745, 3.4046, 1.5084, 3.2208], + device='cuda:1'), covar=tensor([0.0211, 0.1479, 0.1844, 0.0183, 0.3778, 0.0244, 0.1263, 0.0495], + device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0249, 0.0264, 0.0208, 0.0253, 0.0214, 0.0231, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:36:53,020 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 08:36:53,587 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103507.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:36:54,663 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9836, 2.2389, 1.8096, 2.4965, 2.5857, 2.2086, 2.3964, 2.2152], + device='cuda:1'), covar=tensor([0.2086, 0.1130, 0.3721, 0.0787, 0.0347, 0.0263, 0.0356, 0.0371], + device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0230, 0.0248, 0.0257, 0.0197, 0.0194, 0.0215, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 08:36:57,458 INFO [train.py:901] (1/2) Epoch 37, batch 1850, loss[loss=0.1338, simple_loss=0.2182, pruned_loss=0.02467, over 7291.00 frames. ], tot_loss[loss=0.1319, simple_loss=0.2126, pruned_loss=0.02559, over 1440874.67 frames. ], batch size: 57, lr: 4.45e-03, grad_scale: 8.0 +2023-03-21 08:37:02,911 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 08:37:11,708 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103543.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:37:14,077 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.267e+02 1.720e+02 2.032e+02 2.379e+02 3.358e+02, threshold=4.063e+02, percent-clipped=0.0 +2023-03-21 08:37:17,660 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=103555.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:37:19,842 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9424, 3.6597, 3.6016, 3.6896, 3.3629, 3.2193, 4.0298, 2.6408], + device='cuda:1'), covar=tensor([0.0525, 0.0587, 0.0616, 0.0701, 0.0941, 0.1244, 0.0697, 0.2684], + device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0341, 0.0274, 0.0358, 0.0295, 0.0291, 0.0348, 0.0252], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 08:37:20,694 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 08:37:22,707 INFO [train.py:901] (1/2) Epoch 37, batch 1900, loss[loss=0.1242, simple_loss=0.2061, pruned_loss=0.02112, over 7237.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.2126, pruned_loss=0.02549, over 1441644.53 frames. ], batch size: 55, lr: 4.45e-03, grad_scale: 8.0 +2023-03-21 08:37:37,079 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=103591.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:37:41,177 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2414, 4.3421, 4.1054, 4.3399, 4.1005, 4.2557, 4.6012, 4.6186], + device='cuda:1'), covar=tensor([0.0201, 0.0132, 0.0191, 0.0139, 0.0304, 0.0271, 0.0201, 0.0168], + device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0125, 0.0118, 0.0121, 0.0112, 0.0102, 0.0098, 0.0099], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:37:44,673 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 +2023-03-21 08:37:47,920 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 08:37:49,378 INFO [train.py:901] (1/2) Epoch 37, batch 1950, loss[loss=0.1415, simple_loss=0.2262, pruned_loss=0.02841, over 7314.00 frames. ], tot_loss[loss=0.1315, simple_loss=0.2123, pruned_loss=0.0254, over 1442301.42 frames. ], batch size: 83, lr: 4.45e-03, grad_scale: 8.0 +2023-03-21 08:37:51,570 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7160, 2.5744, 2.5535, 3.7043, 2.0245, 3.7015, 1.6291, 3.3451], + device='cuda:1'), covar=tensor([0.0165, 0.1357, 0.1789, 0.0174, 0.3925, 0.0246, 0.1198, 0.0351], + device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0248, 0.0265, 0.0208, 0.0252, 0.0214, 0.0231, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:37:51,591 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3693, 3.3503, 2.4532, 3.8866, 3.1741, 3.3248, 1.7372, 2.5835], + device='cuda:1'), covar=tensor([0.0477, 0.0733, 0.2478, 0.0636, 0.0444, 0.0481, 0.3934, 0.1805], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0258, 0.0281, 0.0268, 0.0271, 0.0265, 0.0235, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:37:58,978 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 08:38:03,647 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 08:38:04,207 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 08:38:06,184 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.676e+02 1.908e+02 2.226e+02 3.747e+02, threshold=3.816e+02, percent-clipped=0.0 +2023-03-21 08:38:14,818 INFO [train.py:901] (1/2) Epoch 37, batch 2000, loss[loss=0.1267, simple_loss=0.2117, pruned_loss=0.02082, over 7278.00 frames. ], tot_loss[loss=0.1313, simple_loss=0.2122, pruned_loss=0.02518, over 1441348.02 frames. ], batch size: 66, lr: 4.45e-03, grad_scale: 8.0 +2023-03-21 08:38:21,948 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 08:38:32,183 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 08:38:40,184 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 08:38:41,176 INFO [train.py:901] (1/2) Epoch 37, batch 2050, loss[loss=0.1347, simple_loss=0.2108, pruned_loss=0.02925, over 7263.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.213, pruned_loss=0.02527, over 1442725.94 frames. ], batch size: 52, lr: 4.45e-03, grad_scale: 8.0 +2023-03-21 08:38:44,745 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103722.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 08:38:53,881 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103740.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:38:58,272 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.791e+02 2.137e+02 2.523e+02 4.759e+02, threshold=4.274e+02, percent-clipped=4.0 +2023-03-21 08:39:01,469 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103754.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:39:06,901 INFO [train.py:901] (1/2) Epoch 37, batch 2100, loss[loss=0.1251, simple_loss=0.2063, pruned_loss=0.02201, over 7302.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.2129, pruned_loss=0.02535, over 1441615.10 frames. ], batch size: 49, lr: 4.45e-03, grad_scale: 8.0 +2023-03-21 08:39:07,051 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103765.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:39:14,021 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 08:39:15,158 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.3402, 4.8781, 4.6769, 5.3614, 5.1270, 5.2986, 4.6557, 4.9292], + device='cuda:1'), covar=tensor([0.0781, 0.2334, 0.2025, 0.1019, 0.0838, 0.1078, 0.0814, 0.1033], + device='cuda:1'), in_proj_covar=tensor([0.0257, 0.0383, 0.0293, 0.0306, 0.0224, 0.0364, 0.0225, 0.0272], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:39:16,672 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 08:39:19,255 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103788.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:39:26,152 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103801.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:39:26,575 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=103802.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:39:33,026 INFO [train.py:901] (1/2) Epoch 37, batch 2150, loss[loss=0.1556, simple_loss=0.2343, pruned_loss=0.03848, over 7211.00 frames. ], tot_loss[loss=0.1317, simple_loss=0.2129, pruned_loss=0.02529, over 1443126.07 frames. ], batch size: 93, lr: 4.45e-03, grad_scale: 8.0 +2023-03-21 08:39:38,510 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103826.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:39:43,983 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=103836.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:39:45,052 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9145, 3.7852, 3.0472, 3.4714, 2.8720, 2.3424, 1.8384, 3.9635], + device='cuda:1'), covar=tensor([0.0048, 0.0059, 0.0151, 0.0070, 0.0194, 0.0540, 0.0660, 0.0049], + device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0093, 0.0112, 0.0093, 0.0129, 0.0135, 0.0130, 0.0103], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 08:39:49,835 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.233e+02 1.798e+02 2.078e+02 2.587e+02 4.516e+02, threshold=4.155e+02, percent-clipped=1.0 +2023-03-21 08:39:58,863 INFO [train.py:901] (1/2) Epoch 37, batch 2200, loss[loss=0.1262, simple_loss=0.2125, pruned_loss=0.01994, over 7257.00 frames. ], tot_loss[loss=0.1321, simple_loss=0.2131, pruned_loss=0.02559, over 1444884.54 frames. ], batch size: 89, lr: 4.45e-03, grad_scale: 8.0 +2023-03-21 08:40:02,462 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 08:40:03,819 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 +2023-03-21 08:40:04,647 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103876.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:40:17,715 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1674, 3.7236, 4.2658, 4.2462, 4.3327, 4.2834, 4.3701, 4.2253], + device='cuda:1'), covar=tensor([0.0027, 0.0084, 0.0025, 0.0028, 0.0025, 0.0028, 0.0022, 0.0035], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0069, 0.0057, 0.0055, 0.0054, 0.0059, 0.0048, 0.0076], + device='cuda:1'), out_proj_covar=tensor([8.1656e-05, 1.4067e-04, 1.0427e-04, 9.7282e-05, 9.3241e-05, 1.0437e-04, + 9.2584e-05, 1.4213e-04], device='cuda:1') +2023-03-21 08:40:24,110 INFO [train.py:901] (1/2) Epoch 37, batch 2250, loss[loss=0.1312, simple_loss=0.2142, pruned_loss=0.02409, over 7259.00 frames. ], tot_loss[loss=0.1323, simple_loss=0.2135, pruned_loss=0.02554, over 1444727.31 frames. ], batch size: 89, lr: 4.44e-03, grad_scale: 8.0 +2023-03-21 08:40:26,726 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1489, 2.6673, 3.2413, 3.1019, 3.2460, 3.0062, 2.7201, 3.2673], + device='cuda:1'), covar=tensor([0.0954, 0.0766, 0.0950, 0.1199, 0.0657, 0.1111, 0.1682, 0.1163], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0069, 0.0051, 0.0050, 0.0051, 0.0050, 0.0069, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 08:40:36,116 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103937.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 08:40:36,477 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 08:40:36,984 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 08:40:42,110 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.782e+02 2.116e+02 2.591e+02 5.653e+02, threshold=4.232e+02, percent-clipped=4.0 +2023-03-21 08:40:45,850 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8982, 2.9743, 2.1662, 3.4993, 2.5036, 2.9998, 1.4600, 2.2601], + device='cuda:1'), covar=tensor([0.0393, 0.0803, 0.2509, 0.0721, 0.0499, 0.0589, 0.3656, 0.1721], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0258, 0.0281, 0.0268, 0.0271, 0.0264, 0.0236, 0.0261], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:40:49,146 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 08:40:50,620 INFO [train.py:901] (1/2) Epoch 37, batch 2300, loss[loss=0.1425, simple_loss=0.2307, pruned_loss=0.02714, over 7117.00 frames. ], tot_loss[loss=0.1323, simple_loss=0.2139, pruned_loss=0.02539, over 1443422.10 frames. ], batch size: 98, lr: 4.44e-03, grad_scale: 8.0 +2023-03-21 08:41:03,306 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103990.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:41:06,786 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103997.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:41:12,693 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-21 08:41:20,369 INFO [train.py:901] (1/2) Epoch 37, batch 2350, loss[loss=0.1184, simple_loss=0.1952, pruned_loss=0.02076, over 7144.00 frames. ], tot_loss[loss=0.1321, simple_loss=0.2138, pruned_loss=0.0252, over 1443971.76 frames. ], batch size: 41, lr: 4.44e-03, grad_scale: 8.0 +2023-03-21 08:41:23,027 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3223, 2.5999, 2.5560, 2.3190, 2.3385, 2.3980, 2.0978, 1.8563], + device='cuda:1'), covar=tensor([0.0495, 0.0295, 0.0298, 0.0228, 0.0496, 0.0541, 0.0413, 0.0451], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0037, 0.0038, 0.0037, 0.0035, 0.0035, 0.0040, 0.0039], + device='cuda:1'), out_proj_covar=tensor([9.5776e-05, 9.5493e-05, 9.5175e-05, 9.3475e-05, 9.1550e-05, 9.1366e-05, + 9.9152e-05, 9.9705e-05], device='cuda:1') +2023-03-21 08:41:24,036 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104022.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 08:41:26,813 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 +2023-03-21 08:41:30,703 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5893, 2.3011, 2.4993, 3.6426, 2.0228, 3.5523, 1.5582, 3.3548], + device='cuda:1'), covar=tensor([0.0202, 0.1441, 0.1832, 0.0230, 0.3575, 0.0327, 0.1263, 0.0338], + device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0246, 0.0264, 0.0208, 0.0250, 0.0214, 0.0229, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:41:37,441 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.348e+02 1.773e+02 2.094e+02 2.536e+02 5.250e+02, threshold=4.187e+02, percent-clipped=1.0 +2023-03-21 08:41:38,119 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104049.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:41:39,135 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104051.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:41:39,498 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 08:41:42,782 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104058.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:41:45,573 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 08:41:46,060 INFO [train.py:901] (1/2) Epoch 37, batch 2400, loss[loss=0.1355, simple_loss=0.2232, pruned_loss=0.02393, over 7326.00 frames. ], tot_loss[loss=0.1326, simple_loss=0.2143, pruned_loss=0.02542, over 1445246.75 frames. ], batch size: 75, lr: 4.44e-03, grad_scale: 8.0 +2023-03-21 08:41:48,684 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=104070.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 08:41:55,691 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 08:41:59,385 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 08:42:02,544 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104096.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:42:03,652 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0308, 2.1781, 2.3397, 2.1084, 2.1850, 2.1746, 1.8442, 1.6056], + device='cuda:1'), covar=tensor([0.0376, 0.0396, 0.0174, 0.0195, 0.0354, 0.0322, 0.0456, 0.0400], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0037, 0.0038, 0.0037, 0.0035, 0.0036, 0.0040, 0.0039], + device='cuda:1'), out_proj_covar=tensor([9.6154e-05, 9.6016e-05, 9.5554e-05, 9.3653e-05, 9.2163e-05, 9.1955e-05, + 9.9320e-05, 1.0001e-04], device='cuda:1') +2023-03-21 08:42:09,491 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104110.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:42:12,504 INFO [train.py:901] (1/2) Epoch 37, batch 2450, loss[loss=0.1309, simple_loss=0.2161, pruned_loss=0.02287, over 7289.00 frames. ], tot_loss[loss=0.1323, simple_loss=0.2139, pruned_loss=0.02535, over 1443533.47 frames. ], batch size: 68, lr: 4.44e-03, grad_scale: 8.0 +2023-03-21 08:42:15,665 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104121.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:42:24,601 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 08:42:28,488 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 +2023-03-21 08:42:29,123 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.742e+02 2.006e+02 2.386e+02 4.524e+02, threshold=4.012e+02, percent-clipped=1.0 +2023-03-21 08:42:37,637 INFO [train.py:901] (1/2) Epoch 37, batch 2500, loss[loss=0.1048, simple_loss=0.1855, pruned_loss=0.01211, over 7173.00 frames. ], tot_loss[loss=0.1325, simple_loss=0.2141, pruned_loss=0.02549, over 1445084.92 frames. ], batch size: 39, lr: 4.44e-03, grad_scale: 8.0 +2023-03-21 08:42:40,909 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8557, 3.2714, 2.7674, 3.0715, 3.0882, 2.5733, 3.1194, 2.9384], + device='cuda:1'), covar=tensor([0.0665, 0.0544, 0.0839, 0.0724, 0.0854, 0.0966, 0.0660, 0.1101], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0057, 0.0065, 0.0058, 0.0056, 0.0060, 0.0055, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:42:49,424 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 08:43:04,292 INFO [train.py:901] (1/2) Epoch 37, batch 2550, loss[loss=0.115, simple_loss=0.2027, pruned_loss=0.01365, over 7322.00 frames. ], tot_loss[loss=0.133, simple_loss=0.2145, pruned_loss=0.02578, over 1444110.78 frames. ], batch size: 44, lr: 4.44e-03, grad_scale: 8.0 +2023-03-21 08:43:12,859 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104232.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 08:43:20,884 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.265e+02 1.707e+02 2.054e+02 2.432e+02 3.967e+02, threshold=4.108e+02, percent-clipped=0.0 +2023-03-21 08:43:29,872 INFO [train.py:901] (1/2) Epoch 37, batch 2600, loss[loss=0.1181, simple_loss=0.2018, pruned_loss=0.01723, over 7275.00 frames. ], tot_loss[loss=0.1329, simple_loss=0.2142, pruned_loss=0.02575, over 1444159.78 frames. ], batch size: 77, lr: 4.44e-03, grad_scale: 8.0 +2023-03-21 08:43:54,441 INFO [train.py:901] (1/2) Epoch 37, batch 2650, loss[loss=0.1429, simple_loss=0.2296, pruned_loss=0.02809, over 7299.00 frames. ], tot_loss[loss=0.1326, simple_loss=0.2141, pruned_loss=0.02556, over 1445056.19 frames. ], batch size: 68, lr: 4.44e-03, grad_scale: 8.0 +2023-03-21 08:44:03,461 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6103, 3.6617, 3.0126, 4.0383, 3.3913, 3.8993, 1.9299, 3.0921], + device='cuda:1'), covar=tensor([0.0484, 0.0689, 0.1836, 0.0457, 0.0474, 0.0685, 0.3253, 0.1616], + device='cuda:1'), in_proj_covar=tensor([0.0280, 0.0259, 0.0284, 0.0270, 0.0273, 0.0267, 0.0238, 0.0261], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:44:03,614 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 +2023-03-21 08:44:09,760 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104346.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:44:10,636 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 1.763e+02 1.995e+02 2.482e+02 3.985e+02, threshold=3.989e+02, percent-clipped=0.0 +2023-03-21 08:44:13,169 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104353.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:44:19,669 INFO [train.py:901] (1/2) Epoch 37, batch 2700, loss[loss=0.137, simple_loss=0.2214, pruned_loss=0.02634, over 7247.00 frames. ], tot_loss[loss=0.1321, simple_loss=0.2133, pruned_loss=0.0254, over 1441030.03 frames. ], batch size: 55, lr: 4.43e-03, grad_scale: 8.0 +2023-03-21 08:44:19,843 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0266, 2.4411, 1.7831, 2.6893, 2.6370, 2.7571, 2.5650, 2.5963], + device='cuda:1'), covar=tensor([0.2196, 0.1027, 0.4126, 0.0943, 0.0333, 0.0311, 0.0431, 0.0445], + device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0228, 0.0246, 0.0255, 0.0195, 0.0192, 0.0214, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 08:44:27,779 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5704, 3.7233, 3.5592, 3.6905, 3.3615, 3.6826, 3.9132, 3.9541], + device='cuda:1'), covar=tensor([0.0250, 0.0184, 0.0233, 0.0193, 0.0428, 0.0369, 0.0311, 0.0230], + device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0125, 0.0118, 0.0122, 0.0113, 0.0102, 0.0099, 0.0099], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:44:30,304 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8799, 3.8111, 3.1832, 3.5454, 2.8192, 2.2919, 1.7608, 4.0052], + device='cuda:1'), covar=tensor([0.0056, 0.0058, 0.0140, 0.0075, 0.0189, 0.0580, 0.0713, 0.0050], + device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0094, 0.0114, 0.0094, 0.0129, 0.0137, 0.0131, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 08:44:35,242 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104396.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:44:39,933 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104405.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:44:40,267 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 +2023-03-21 08:44:40,423 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.5023, 5.1039, 4.8435, 5.4971, 5.2636, 5.4574, 4.7576, 5.1139], + device='cuda:1'), covar=tensor([0.0649, 0.2233, 0.2141, 0.0912, 0.0955, 0.0964, 0.0733, 0.0995], + device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0382, 0.0294, 0.0307, 0.0224, 0.0363, 0.0223, 0.0272], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:44:44,847 INFO [train.py:901] (1/2) Epoch 37, batch 2750, loss[loss=0.1252, simple_loss=0.2132, pruned_loss=0.0186, over 7230.00 frames. ], tot_loss[loss=0.1319, simple_loss=0.2134, pruned_loss=0.0252, over 1441992.97 frames. ], batch size: 93, lr: 4.43e-03, grad_scale: 8.0 +2023-03-21 08:44:47,898 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104421.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:44:59,167 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=104444.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:45:01,046 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.660e+02 2.023e+02 2.410e+02 4.270e+02, threshold=4.046e+02, percent-clipped=1.0 +2023-03-21 08:45:02,342 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.86 vs. limit=5.0 +2023-03-21 08:45:09,297 INFO [train.py:901] (1/2) Epoch 37, batch 2800, loss[loss=0.1415, simple_loss=0.2196, pruned_loss=0.03175, over 7247.00 frames. ], tot_loss[loss=0.1316, simple_loss=0.2127, pruned_loss=0.02524, over 1438954.75 frames. ], batch size: 55, lr: 4.43e-03, grad_scale: 8.0 +2023-03-21 08:45:11,300 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=104469.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:45:34,907 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 08:45:34,958 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 08:45:34,983 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 08:45:34,998 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 08:45:35,477 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 08:45:35,493 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 08:45:35,503 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 08:45:35,599 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 08:45:35,676 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 08:45:35,691 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 08:45:35,707 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 08:45:35,872 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 08:45:35,878 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 08:45:35,885 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 08:45:36,026 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 08:45:36,029 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 08:45:36,087 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 08:45:40,894 INFO [train.py:901] (1/2) Epoch 38, batch 0, loss[loss=0.1485, simple_loss=0.2233, pruned_loss=0.0368, over 7334.00 frames. ], tot_loss[loss=0.1485, simple_loss=0.2233, pruned_loss=0.0368, over 7334.00 frames. ], batch size: 54, lr: 4.37e-03, grad_scale: 8.0 +2023-03-21 08:45:40,895 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 08:46:03,636 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9226, 1.9110, 2.1795, 2.4380, 2.2086, 2.4193, 2.3835, 2.5146], + device='cuda:1'), covar=tensor([0.3272, 0.3250, 0.2244, 0.1815, 0.2220, 0.1834, 0.2562, 0.2724], + device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0077, 0.0070, 0.0062, 0.0062, 0.0061, 0.0102, 0.0065], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 08:46:07,744 INFO [train.py:935] (1/2) Epoch 38, validation: loss=0.1635, simple_loss=0.256, pruned_loss=0.03545, over 1622729.00 frames. +2023-03-21 08:46:07,745 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 08:46:10,023 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-03-21 08:46:15,255 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 08:46:18,636 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-03-21 08:46:25,915 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 08:46:29,969 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104532.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 08:46:32,942 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 08:46:33,434 INFO [train.py:901] (1/2) Epoch 38, batch 50, loss[loss=0.1257, simple_loss=0.2177, pruned_loss=0.01684, over 7122.00 frames. ], tot_loss[loss=0.1328, simple_loss=0.214, pruned_loss=0.02577, over 323434.40 frames. ], batch size: 98, lr: 4.37e-03, grad_scale: 8.0 +2023-03-21 08:46:34,429 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 08:46:36,899 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 08:46:37,844 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.288e+02 1.780e+02 2.077e+02 2.791e+02 4.697e+02, threshold=4.154e+02, percent-clipped=2.0 +2023-03-21 08:46:53,971 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 08:46:54,465 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 08:46:54,517 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=104580.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:46:59,015 INFO [train.py:901] (1/2) Epoch 38, batch 100, loss[loss=0.1217, simple_loss=0.2054, pruned_loss=0.01899, over 7325.00 frames. ], tot_loss[loss=0.1327, simple_loss=0.2144, pruned_loss=0.02552, over 571741.11 frames. ], batch size: 44, lr: 4.37e-03, grad_scale: 8.0 +2023-03-21 08:47:13,109 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 +2023-03-21 08:47:22,579 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3936, 3.3827, 2.9878, 3.3986, 3.1795, 2.9028, 3.3005, 2.9751], + device='cuda:1'), covar=tensor([0.0390, 0.0790, 0.1229, 0.1348, 0.1812, 0.1114, 0.0529, 0.1258], + device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0057, 0.0065, 0.0058, 0.0055, 0.0060, 0.0055, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:47:24,972 INFO [train.py:901] (1/2) Epoch 38, batch 150, loss[loss=0.1425, simple_loss=0.2229, pruned_loss=0.03108, over 7272.00 frames. ], tot_loss[loss=0.1335, simple_loss=0.2142, pruned_loss=0.02643, over 764865.83 frames. ], batch size: 77, lr: 4.37e-03, grad_scale: 8.0 +2023-03-21 08:47:28,538 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104646.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:47:29,416 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.300e+02 1.726e+02 2.045e+02 2.303e+02 3.131e+02, threshold=4.089e+02, percent-clipped=0.0 +2023-03-21 08:47:32,736 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104653.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:47:50,591 INFO [train.py:901] (1/2) Epoch 38, batch 200, loss[loss=0.1265, simple_loss=0.2107, pruned_loss=0.0212, over 7270.00 frames. ], tot_loss[loss=0.1328, simple_loss=0.2132, pruned_loss=0.02613, over 916880.63 frames. ], batch size: 52, lr: 4.37e-03, grad_scale: 8.0 +2023-03-21 08:47:53,706 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=104694.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:47:56,181 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 08:47:57,246 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=104701.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:47:59,303 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104705.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:48:01,205 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 08:48:08,211 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 08:48:16,325 INFO [train.py:901] (1/2) Epoch 38, batch 250, loss[loss=0.1358, simple_loss=0.2094, pruned_loss=0.03116, over 7261.00 frames. ], tot_loss[loss=0.132, simple_loss=0.2129, pruned_loss=0.02558, over 1033803.15 frames. ], batch size: 47, lr: 4.37e-03, grad_scale: 8.0 +2023-03-21 08:48:21,305 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.231e+02 1.611e+02 1.949e+02 2.401e+02 6.980e+02, threshold=3.898e+02, percent-clipped=1.0 +2023-03-21 08:48:21,339 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 08:48:23,906 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=104753.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:48:41,927 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 08:48:42,356 INFO [train.py:901] (1/2) Epoch 38, batch 300, loss[loss=0.1464, simple_loss=0.2275, pruned_loss=0.03269, over 7334.00 frames. ], tot_loss[loss=0.132, simple_loss=0.2132, pruned_loss=0.02535, over 1126058.25 frames. ], batch size: 75, lr: 4.37e-03, grad_scale: 8.0 +2023-03-21 08:48:49,312 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8880, 2.7529, 2.7307, 3.8228, 2.1031, 3.8728, 1.5309, 3.4510], + device='cuda:1'), covar=tensor([0.0216, 0.1252, 0.1712, 0.0175, 0.3984, 0.0247, 0.1241, 0.0476], + device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0246, 0.0260, 0.0205, 0.0249, 0.0212, 0.0227, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:48:50,610 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 08:48:59,095 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2787, 2.5332, 2.5929, 2.2629, 2.7124, 2.3840, 2.1312, 1.9565], + device='cuda:1'), covar=tensor([0.0700, 0.0441, 0.0248, 0.0526, 0.0622, 0.0476, 0.0523, 0.0436], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0038, 0.0038, 0.0038, 0.0035, 0.0036, 0.0040, 0.0040], + device='cuda:1'), out_proj_covar=tensor([9.7466e-05, 9.7867e-05, 9.6852e-05, 9.5469e-05, 9.3222e-05, 9.3352e-05, + 1.0052e-04, 1.0129e-04], device='cuda:1') +2023-03-21 08:49:07,516 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8974, 3.1479, 3.8796, 3.9002, 4.0520, 3.9921, 3.9884, 3.6230], + device='cuda:1'), covar=tensor([0.0049, 0.0168, 0.0052, 0.0047, 0.0046, 0.0042, 0.0057, 0.0085], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0069, 0.0057, 0.0055, 0.0055, 0.0060, 0.0048, 0.0077], + device='cuda:1'), out_proj_covar=tensor([8.2391e-05, 1.4246e-04, 1.0471e-04, 9.6653e-05, 9.3659e-05, 1.0442e-04, + 9.1468e-05, 1.4358e-04], device='cuda:1') +2023-03-21 08:49:07,910 INFO [train.py:901] (1/2) Epoch 38, batch 350, loss[loss=0.1481, simple_loss=0.2222, pruned_loss=0.03693, over 7320.00 frames. ], tot_loss[loss=0.1322, simple_loss=0.213, pruned_loss=0.02574, over 1195586.78 frames. ], batch size: 49, lr: 4.37e-03, grad_scale: 8.0 +2023-03-21 08:49:12,390 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.194e+02 1.765e+02 1.993e+02 2.346e+02 3.883e+02, threshold=3.986e+02, percent-clipped=0.0 +2023-03-21 08:49:25,533 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 08:49:33,213 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9924, 3.3252, 3.9759, 3.9908, 4.0621, 4.0378, 4.1150, 3.9084], + device='cuda:1'), covar=tensor([0.0029, 0.0115, 0.0029, 0.0024, 0.0028, 0.0027, 0.0030, 0.0047], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0069, 0.0057, 0.0055, 0.0055, 0.0060, 0.0048, 0.0077], + device='cuda:1'), out_proj_covar=tensor([8.2231e-05, 1.4227e-04, 1.0441e-04, 9.6767e-05, 9.3763e-05, 1.0463e-04, + 9.1309e-05, 1.4347e-04], device='cuda:1') +2023-03-21 08:49:33,600 INFO [train.py:901] (1/2) Epoch 38, batch 400, loss[loss=0.1276, simple_loss=0.2147, pruned_loss=0.02028, over 7253.00 frames. ], tot_loss[loss=0.1322, simple_loss=0.213, pruned_loss=0.02566, over 1251468.46 frames. ], batch size: 52, lr: 4.37e-03, grad_scale: 8.0 +2023-03-21 08:49:35,860 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 08:49:48,044 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-03-21 08:49:59,217 INFO [train.py:901] (1/2) Epoch 38, batch 450, loss[loss=0.139, simple_loss=0.2243, pruned_loss=0.02685, over 7268.00 frames. ], tot_loss[loss=0.1327, simple_loss=0.2134, pruned_loss=0.02598, over 1294770.05 frames. ], batch size: 52, lr: 4.36e-03, grad_scale: 8.0 +2023-03-21 08:50:03,664 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.059e+02 1.744e+02 1.979e+02 2.437e+02 3.979e+02, threshold=3.958e+02, percent-clipped=0.0 +2023-03-21 08:50:05,674 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 08:50:06,210 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 08:50:24,969 INFO [train.py:901] (1/2) Epoch 38, batch 500, loss[loss=0.1367, simple_loss=0.2208, pruned_loss=0.02625, over 7311.00 frames. ], tot_loss[loss=0.1332, simple_loss=0.214, pruned_loss=0.02613, over 1328572.24 frames. ], batch size: 80, lr: 4.36e-03, grad_scale: 8.0 +2023-03-21 08:50:27,767 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-03-21 08:50:28,546 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9763, 3.8061, 3.6490, 3.8054, 3.1842, 3.5335, 3.9148, 3.4349], + device='cuda:1'), covar=tensor([0.0300, 0.0227, 0.0225, 0.0225, 0.0984, 0.0236, 0.0268, 0.0289], + device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0101, 0.0104, 0.0089, 0.0177, 0.0107, 0.0104, 0.0112], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:50:38,899 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 08:50:39,884 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 08:50:40,933 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 08:50:42,945 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 08:50:47,501 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 08:50:51,554 INFO [train.py:901] (1/2) Epoch 38, batch 550, loss[loss=0.1507, simple_loss=0.2304, pruned_loss=0.03547, over 7264.00 frames. ], tot_loss[loss=0.1331, simple_loss=0.2137, pruned_loss=0.02621, over 1352815.10 frames. ], batch size: 89, lr: 4.36e-03, grad_scale: 8.0 +2023-03-21 08:50:56,112 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.231e+02 1.699e+02 2.007e+02 2.448e+02 5.242e+02, threshold=4.013e+02, percent-clipped=3.0 +2023-03-21 08:50:59,104 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 08:51:04,715 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7605, 4.3222, 4.1132, 4.7599, 4.5481, 4.6737, 4.0452, 4.3298], + device='cuda:1'), covar=tensor([0.0951, 0.2417, 0.2645, 0.0995, 0.0917, 0.1212, 0.0848, 0.1215], + device='cuda:1'), in_proj_covar=tensor([0.0259, 0.0386, 0.0298, 0.0310, 0.0225, 0.0366, 0.0226, 0.0273], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:51:07,132 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 08:51:10,663 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 08:51:17,242 INFO [train.py:901] (1/2) Epoch 38, batch 600, loss[loss=0.1427, simple_loss=0.2246, pruned_loss=0.03038, over 7330.00 frames. ], tot_loss[loss=0.1325, simple_loss=0.2133, pruned_loss=0.02587, over 1374069.67 frames. ], batch size: 59, lr: 4.36e-03, grad_scale: 8.0 +2023-03-21 08:51:17,764 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 08:51:32,366 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9642, 2.4727, 3.0262, 2.8793, 3.0954, 2.7622, 2.5006, 3.0739], + device='cuda:1'), covar=tensor([0.1258, 0.0818, 0.1092, 0.1185, 0.0680, 0.1191, 0.1923, 0.1215], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0067, 0.0050, 0.0050, 0.0050, 0.0048, 0.0067, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 08:51:33,247 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 08:51:41,394 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6557, 4.1200, 4.1540, 4.3564, 4.2901, 4.1705, 4.5598, 3.8651], + device='cuda:1'), covar=tensor([0.0139, 0.0193, 0.0147, 0.0146, 0.0448, 0.0132, 0.0128, 0.0236], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0102, 0.0104, 0.0089, 0.0179, 0.0108, 0.0105, 0.0114], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:51:42,765 INFO [train.py:901] (1/2) Epoch 38, batch 650, loss[loss=0.1085, simple_loss=0.1698, pruned_loss=0.02363, over 5932.00 frames. ], tot_loss[loss=0.1315, simple_loss=0.2126, pruned_loss=0.02523, over 1386692.37 frames. ], batch size: 25, lr: 4.36e-03, grad_scale: 8.0 +2023-03-21 08:51:42,766 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 08:51:47,278 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.200e+02 1.709e+02 2.034e+02 2.305e+02 4.227e+02, threshold=4.069e+02, percent-clipped=1.0 +2023-03-21 08:51:47,375 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.3869, 4.8516, 4.7925, 5.3056, 5.1875, 5.2455, 4.4119, 4.9325], + device='cuda:1'), covar=tensor([0.0683, 0.2347, 0.2037, 0.0942, 0.0825, 0.1184, 0.0810, 0.1065], + device='cuda:1'), in_proj_covar=tensor([0.0261, 0.0389, 0.0299, 0.0313, 0.0226, 0.0368, 0.0228, 0.0275], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:51:59,363 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 08:52:08,284 INFO [train.py:901] (1/2) Epoch 38, batch 700, loss[loss=0.1246, simple_loss=0.203, pruned_loss=0.02313, over 7199.00 frames. ], tot_loss[loss=0.1314, simple_loss=0.2127, pruned_loss=0.02511, over 1399390.36 frames. ], batch size: 50, lr: 4.36e-03, grad_scale: 8.0 +2023-03-21 08:52:08,813 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 08:52:31,138 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-03-21 08:52:33,679 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 08:52:33,694 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 08:52:34,239 INFO [train.py:901] (1/2) Epoch 38, batch 750, loss[loss=0.1046, simple_loss=0.1737, pruned_loss=0.0177, over 7022.00 frames. ], tot_loss[loss=0.1314, simple_loss=0.2124, pruned_loss=0.02524, over 1407714.97 frames. ], batch size: 35, lr: 4.36e-03, grad_scale: 8.0 +2023-03-21 08:52:38,657 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.214e+02 1.746e+02 1.996e+02 2.385e+02 4.034e+02, threshold=3.992e+02, percent-clipped=0.0 +2023-03-21 08:52:47,024 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0976, 2.5014, 1.7888, 2.9460, 2.9125, 2.8706, 2.5000, 2.5576], + device='cuda:1'), covar=tensor([0.2198, 0.0951, 0.3835, 0.0748, 0.0352, 0.0367, 0.0382, 0.0431], + device='cuda:1'), in_proj_covar=tensor([0.0254, 0.0232, 0.0251, 0.0259, 0.0199, 0.0195, 0.0216, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:52:48,831 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 08:52:52,821 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 08:52:59,301 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 08:52:59,739 INFO [train.py:901] (1/2) Epoch 38, batch 800, loss[loss=0.156, simple_loss=0.2319, pruned_loss=0.04003, over 7344.00 frames. ], tot_loss[loss=0.1312, simple_loss=0.212, pruned_loss=0.02513, over 1412413.82 frames. ], batch size: 54, lr: 4.36e-03, grad_scale: 8.0 +2023-03-21 08:53:00,744 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 08:53:11,778 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 08:53:25,509 INFO [train.py:901] (1/2) Epoch 38, batch 850, loss[loss=0.1263, simple_loss=0.2114, pruned_loss=0.02064, over 7273.00 frames. ], tot_loss[loss=0.1311, simple_loss=0.2119, pruned_loss=0.02517, over 1417557.13 frames. ], batch size: 66, lr: 4.36e-03, grad_scale: 8.0 +2023-03-21 08:53:26,348 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-03-21 08:53:30,015 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.181e+02 1.670e+02 2.029e+02 2.408e+02 3.853e+02, threshold=4.058e+02, percent-clipped=0.0 +2023-03-21 08:53:31,054 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 08:53:31,063 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 08:53:31,125 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.7455, 5.3555, 5.1583, 5.6881, 5.5000, 5.6578, 5.1047, 5.3610], + device='cuda:1'), covar=tensor([0.0691, 0.1744, 0.1905, 0.0862, 0.0953, 0.1030, 0.0655, 0.1046], + device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0393, 0.0303, 0.0316, 0.0230, 0.0372, 0.0230, 0.0278], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:53:37,228 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 08:53:40,780 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 08:53:50,853 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105388.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:53:51,206 INFO [train.py:901] (1/2) Epoch 38, batch 900, loss[loss=0.1254, simple_loss=0.2108, pruned_loss=0.02, over 7343.00 frames. ], tot_loss[loss=0.1307, simple_loss=0.2117, pruned_loss=0.02485, over 1422526.64 frames. ], batch size: 73, lr: 4.35e-03, grad_scale: 8.0 +2023-03-21 08:54:07,715 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0212, 4.5344, 4.3761, 4.9506, 4.7760, 4.8796, 4.2614, 4.5967], + device='cuda:1'), covar=tensor([0.0744, 0.2172, 0.2145, 0.0864, 0.0871, 0.1108, 0.0864, 0.1069], + device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0392, 0.0300, 0.0313, 0.0229, 0.0370, 0.0228, 0.0276], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 08:54:17,257 INFO [train.py:901] (1/2) Epoch 38, batch 950, loss[loss=0.1304, simple_loss=0.2088, pruned_loss=0.02596, over 7276.00 frames. ], tot_loss[loss=0.1308, simple_loss=0.2122, pruned_loss=0.02472, over 1428771.21 frames. ], batch size: 52, lr: 4.35e-03, grad_scale: 16.0 +2023-03-21 08:54:18,272 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 08:54:21,246 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-03-21 08:54:22,332 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.229e+02 1.656e+02 1.886e+02 2.311e+02 4.194e+02, threshold=3.772e+02, percent-clipped=1.0 +2023-03-21 08:54:23,023 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105449.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 08:54:29,242 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 +2023-03-21 08:54:41,205 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 08:54:42,326 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105486.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:54:43,669 INFO [train.py:901] (1/2) Epoch 38, batch 1000, loss[loss=0.1521, simple_loss=0.23, pruned_loss=0.03706, over 7330.00 frames. ], tot_loss[loss=0.1309, simple_loss=0.2124, pruned_loss=0.02467, over 1432446.87 frames. ], batch size: 61, lr: 4.35e-03, grad_scale: 16.0 +2023-03-21 08:54:47,007 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 +2023-03-21 08:54:50,279 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105502.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:55:00,896 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 08:55:09,310 INFO [train.py:901] (1/2) Epoch 38, batch 1050, loss[loss=0.1264, simple_loss=0.209, pruned_loss=0.02188, over 7325.00 frames. ], tot_loss[loss=0.1309, simple_loss=0.212, pruned_loss=0.02484, over 1433095.67 frames. ], batch size: 80, lr: 4.35e-03, grad_scale: 16.0 +2023-03-21 08:55:13,467 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105547.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:55:13,791 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.256e+02 1.762e+02 2.024e+02 2.467e+02 4.563e+02, threshold=4.048e+02, percent-clipped=5.0 +2023-03-21 08:55:21,776 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 08:55:21,922 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105563.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:55:25,877 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 08:55:34,884 INFO [train.py:901] (1/2) Epoch 38, batch 1100, loss[loss=0.1225, simple_loss=0.2109, pruned_loss=0.01709, over 7275.00 frames. ], tot_loss[loss=0.1313, simple_loss=0.2123, pruned_loss=0.0251, over 1434847.36 frames. ], batch size: 52, lr: 4.35e-03, grad_scale: 16.0 +2023-03-21 08:55:56,056 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 08:55:56,068 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 08:56:01,019 INFO [train.py:901] (1/2) Epoch 38, batch 1150, loss[loss=0.1334, simple_loss=0.2177, pruned_loss=0.02453, over 7212.00 frames. ], tot_loss[loss=0.1311, simple_loss=0.2119, pruned_loss=0.02513, over 1436361.23 frames. ], batch size: 93, lr: 4.35e-03, grad_scale: 16.0 +2023-03-21 08:56:05,512 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.238e+02 1.823e+02 2.073e+02 2.521e+02 5.443e+02, threshold=4.146e+02, percent-clipped=2.0 +2023-03-21 08:56:09,177 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 08:56:09,656 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 08:56:11,529 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 +2023-03-21 08:56:26,771 INFO [train.py:901] (1/2) Epoch 38, batch 1200, loss[loss=0.1393, simple_loss=0.2202, pruned_loss=0.02923, over 7274.00 frames. ], tot_loss[loss=0.131, simple_loss=0.2119, pruned_loss=0.02506, over 1434926.70 frames. ], batch size: 52, lr: 4.35e-03, grad_scale: 16.0 +2023-03-21 08:56:43,618 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 08:56:45,546 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 08:56:53,214 INFO [train.py:901] (1/2) Epoch 38, batch 1250, loss[loss=0.1341, simple_loss=0.2244, pruned_loss=0.02193, over 7120.00 frames. ], tot_loss[loss=0.1305, simple_loss=0.2114, pruned_loss=0.02479, over 1436342.08 frames. ], batch size: 98, lr: 4.35e-03, grad_scale: 8.0 +2023-03-21 08:56:53,402 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3748, 3.4274, 2.7341, 3.8697, 3.1551, 3.2292, 1.8421, 2.7027], + device='cuda:1'), covar=tensor([0.0425, 0.0642, 0.2356, 0.0434, 0.0403, 0.0394, 0.3876, 0.1896], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0257, 0.0281, 0.0269, 0.0270, 0.0264, 0.0235, 0.0259], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:56:55,793 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105744.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 08:56:58,134 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.196e+02 1.762e+02 2.046e+02 2.416e+02 3.370e+02, threshold=4.092e+02, percent-clipped=0.0 +2023-03-21 08:57:06,635 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 08:57:11,135 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 08:57:12,639 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 08:57:18,897 INFO [train.py:901] (1/2) Epoch 38, batch 1300, loss[loss=0.1298, simple_loss=0.2153, pruned_loss=0.02218, over 7221.00 frames. ], tot_loss[loss=0.1311, simple_loss=0.2124, pruned_loss=0.02494, over 1439712.82 frames. ], batch size: 93, lr: 4.35e-03, grad_scale: 8.0 +2023-03-21 08:57:20,507 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105792.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:57:24,025 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105799.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:57:32,480 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0874, 3.2818, 2.2927, 3.5651, 2.6710, 3.0525, 1.5334, 2.3719], + device='cuda:1'), covar=tensor([0.0466, 0.0858, 0.2768, 0.0664, 0.0502, 0.0558, 0.3882, 0.2003], + device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0259, 0.0283, 0.0270, 0.0272, 0.0266, 0.0237, 0.0261], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:57:35,426 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2365, 4.1686, 3.8432, 3.8885, 3.4383, 2.5026, 1.9902, 4.3341], + device='cuda:1'), covar=tensor([0.0060, 0.0069, 0.0111, 0.0066, 0.0154, 0.0576, 0.0708, 0.0053], + device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0095, 0.0115, 0.0095, 0.0132, 0.0139, 0.0132, 0.0106], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 08:57:36,928 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 08:57:38,919 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 08:57:43,001 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 08:57:44,961 INFO [train.py:901] (1/2) Epoch 38, batch 1350, loss[loss=0.1389, simple_loss=0.2217, pruned_loss=0.02805, over 7342.00 frames. ], tot_loss[loss=0.1313, simple_loss=0.2126, pruned_loss=0.02506, over 1441849.92 frames. ], batch size: 75, lr: 4.35e-03, grad_scale: 8.0 +2023-03-21 08:57:46,572 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105842.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:57:49,967 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 1.666e+02 2.004e+02 2.449e+02 3.831e+02, threshold=4.008e+02, percent-clipped=0.0 +2023-03-21 08:57:52,092 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105853.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:57:52,959 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 08:57:54,546 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105858.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:57:55,615 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105860.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:57:59,647 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5822, 1.4607, 1.5976, 1.8566, 1.6832, 1.8489, 1.4390, 1.8985], + device='cuda:1'), covar=tensor([0.2042, 0.3434, 0.2211, 0.1345, 0.1534, 0.1307, 0.2426, 0.1548], + device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0080, 0.0072, 0.0064, 0.0062, 0.0061, 0.0104, 0.0067], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 08:58:10,536 INFO [train.py:901] (1/2) Epoch 38, batch 1400, loss[loss=0.1406, simple_loss=0.2174, pruned_loss=0.03195, over 7215.00 frames. ], tot_loss[loss=0.1313, simple_loss=0.2126, pruned_loss=0.025, over 1443445.41 frames. ], batch size: 50, lr: 4.34e-03, grad_scale: 8.0 +2023-03-21 08:58:15,133 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1451, 2.4561, 1.8510, 2.9767, 2.9756, 2.7958, 2.7507, 2.6606], + device='cuda:1'), covar=tensor([0.2280, 0.1109, 0.3908, 0.0774, 0.0268, 0.0255, 0.0438, 0.0368], + device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0233, 0.0252, 0.0261, 0.0201, 0.0197, 0.0219, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:58:25,171 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 08:58:27,028 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.33 vs. limit=5.0 +2023-03-21 08:58:35,972 INFO [train.py:901] (1/2) Epoch 38, batch 1450, loss[loss=0.134, simple_loss=0.2214, pruned_loss=0.02327, over 7333.00 frames. ], tot_loss[loss=0.1319, simple_loss=0.2135, pruned_loss=0.02519, over 1444994.88 frames. ], batch size: 75, lr: 4.34e-03, grad_scale: 8.0 +2023-03-21 08:58:41,018 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.294e+02 1.750e+02 2.032e+02 2.346e+02 3.706e+02, threshold=4.064e+02, percent-clipped=0.0 +2023-03-21 08:58:49,303 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 08:58:51,055 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-03-21 08:58:54,388 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1796, 2.4109, 1.9370, 2.9213, 2.8954, 2.5872, 2.6134, 2.5692], + device='cuda:1'), covar=tensor([0.2147, 0.1137, 0.3809, 0.0592, 0.0329, 0.0277, 0.0474, 0.0479], + device='cuda:1'), in_proj_covar=tensor([0.0256, 0.0232, 0.0252, 0.0260, 0.0201, 0.0196, 0.0218, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:59:00,577 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 +2023-03-21 08:59:01,713 INFO [train.py:901] (1/2) Epoch 38, batch 1500, loss[loss=0.1036, simple_loss=0.1665, pruned_loss=0.02035, over 5985.00 frames. ], tot_loss[loss=0.1311, simple_loss=0.2122, pruned_loss=0.02499, over 1439490.01 frames. ], batch size: 25, lr: 4.34e-03, grad_scale: 8.0 +2023-03-21 08:59:06,404 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 08:59:27,813 INFO [train.py:901] (1/2) Epoch 38, batch 1550, loss[loss=0.1457, simple_loss=0.2181, pruned_loss=0.03662, over 7309.00 frames. ], tot_loss[loss=0.1313, simple_loss=0.2124, pruned_loss=0.02505, over 1441450.68 frames. ], batch size: 49, lr: 4.34e-03, grad_scale: 8.0 +2023-03-21 08:59:29,319 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 08:59:30,507 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106044.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:59:33,487 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.162e+02 1.752e+02 2.054e+02 2.320e+02 6.446e+02, threshold=4.109e+02, percent-clipped=1.0 +2023-03-21 08:59:47,818 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106077.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 08:59:48,356 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2323, 2.5320, 2.0018, 2.8300, 3.0736, 2.5243, 2.5966, 2.6668], + device='cuda:1'), covar=tensor([0.2325, 0.1144, 0.4076, 0.0685, 0.0319, 0.0256, 0.0394, 0.0447], + device='cuda:1'), in_proj_covar=tensor([0.0258, 0.0233, 0.0253, 0.0261, 0.0202, 0.0197, 0.0219, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 08:59:54,272 INFO [train.py:901] (1/2) Epoch 38, batch 1600, loss[loss=0.1418, simple_loss=0.2245, pruned_loss=0.02949, over 7278.00 frames. ], tot_loss[loss=0.1307, simple_loss=0.2117, pruned_loss=0.02482, over 1439311.45 frames. ], batch size: 77, lr: 4.34e-03, grad_scale: 8.0 +2023-03-21 08:59:55,845 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=106092.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:00:02,441 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 09:00:02,966 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 09:00:05,502 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 09:00:15,607 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 09:00:19,938 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106138.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:00:20,302 INFO [train.py:901] (1/2) Epoch 38, batch 1650, loss[loss=0.1369, simple_loss=0.2119, pruned_loss=0.03093, over 7212.00 frames. ], tot_loss[loss=0.1309, simple_loss=0.2117, pruned_loss=0.02504, over 1437868.43 frames. ], batch size: 50, lr: 4.34e-03, grad_scale: 8.0 +2023-03-21 09:00:20,305 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 09:00:21,928 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106142.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:00:24,928 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106148.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:00:25,299 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.144e+02 1.757e+02 2.005e+02 2.321e+02 4.774e+02, threshold=4.009e+02, percent-clipped=1.0 +2023-03-21 09:00:27,847 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 09:00:28,414 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106155.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:00:29,995 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106158.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:00:37,853 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 +2023-03-21 09:00:44,087 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 09:00:45,980 INFO [train.py:901] (1/2) Epoch 38, batch 1700, loss[loss=0.1426, simple_loss=0.2263, pruned_loss=0.02943, over 7375.00 frames. ], tot_loss[loss=0.131, simple_loss=0.2119, pruned_loss=0.02507, over 1435473.93 frames. ], batch size: 73, lr: 4.34e-03, grad_scale: 8.0 +2023-03-21 09:00:46,550 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=106190.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:00:48,579 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 09:00:54,912 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=106206.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:00:59,370 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 09:01:07,183 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7577, 1.4684, 2.0087, 2.3422, 2.0645, 2.0245, 2.0191, 2.3510], + device='cuda:1'), covar=tensor([0.2568, 0.4087, 0.1877, 0.1613, 0.1474, 0.2049, 0.1775, 0.1687], + device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0078, 0.0070, 0.0063, 0.0061, 0.0060, 0.0101, 0.0065], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:01:11,909 INFO [train.py:901] (1/2) Epoch 38, batch 1750, loss[loss=0.1246, simple_loss=0.2106, pruned_loss=0.01932, over 7288.00 frames. ], tot_loss[loss=0.1312, simple_loss=0.2119, pruned_loss=0.02523, over 1436562.43 frames. ], batch size: 66, lr: 4.34e-03, grad_scale: 8.0 +2023-03-21 09:01:16,876 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.200e+02 1.849e+02 2.238e+02 2.695e+02 5.615e+02, threshold=4.475e+02, percent-clipped=4.0 +2023-03-21 09:01:21,110 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7128, 1.5021, 1.9906, 2.2776, 1.9761, 2.0371, 1.9157, 2.2049], + device='cuda:1'), covar=tensor([0.3023, 0.4014, 0.2817, 0.1706, 0.1984, 0.3406, 0.2208, 0.2727], + device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0079, 0.0071, 0.0063, 0.0061, 0.0060, 0.0102, 0.0066], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:01:24,427 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 09:01:25,963 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 09:01:27,076 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106268.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:01:37,318 INFO [train.py:901] (1/2) Epoch 38, batch 1800, loss[loss=0.1555, simple_loss=0.2319, pruned_loss=0.03954, over 7336.00 frames. ], tot_loss[loss=0.1315, simple_loss=0.2126, pruned_loss=0.02525, over 1438576.08 frames. ], batch size: 54, lr: 4.34e-03, grad_scale: 8.0 +2023-03-21 09:01:41,863 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7369, 3.7604, 2.8072, 3.3973, 2.6622, 2.2183, 1.7567, 3.7487], + device='cuda:1'), covar=tensor([0.0057, 0.0052, 0.0191, 0.0086, 0.0205, 0.0592, 0.0701, 0.0066], + device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0094, 0.0115, 0.0095, 0.0131, 0.0137, 0.0132, 0.0105], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 09:01:46,127 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 09:01:57,990 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106329.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:01:59,886 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 09:02:02,092 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106336.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:02:03,438 INFO [train.py:901] (1/2) Epoch 38, batch 1850, loss[loss=0.1201, simple_loss=0.2045, pruned_loss=0.01786, over 7268.00 frames. ], tot_loss[loss=0.132, simple_loss=0.2131, pruned_loss=0.02545, over 1440118.32 frames. ], batch size: 89, lr: 4.34e-03, grad_scale: 8.0 +2023-03-21 09:02:08,336 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.034e+02 1.796e+02 2.169e+02 2.438e+02 6.598e+02, threshold=4.338e+02, percent-clipped=1.0 +2023-03-21 09:02:09,900 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 09:02:26,994 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 09:02:28,467 INFO [train.py:901] (1/2) Epoch 38, batch 1900, loss[loss=0.1361, simple_loss=0.2163, pruned_loss=0.02797, over 7243.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.213, pruned_loss=0.02532, over 1441056.40 frames. ], batch size: 89, lr: 4.33e-03, grad_scale: 8.0 +2023-03-21 09:02:32,676 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106397.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:02:33,770 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2206, 3.9019, 3.8920, 3.8565, 3.8983, 3.7719, 4.0999, 3.6164], + device='cuda:1'), covar=tensor([0.0140, 0.0162, 0.0113, 0.0188, 0.0391, 0.0104, 0.0133, 0.0205], + device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0100, 0.0101, 0.0088, 0.0175, 0.0107, 0.0103, 0.0111], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:02:52,426 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106433.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:02:52,853 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 09:02:55,359 INFO [train.py:901] (1/2) Epoch 38, batch 1950, loss[loss=0.1342, simple_loss=0.2173, pruned_loss=0.02554, over 7307.00 frames. ], tot_loss[loss=0.1313, simple_loss=0.2126, pruned_loss=0.02505, over 1441985.09 frames. ], batch size: 80, lr: 4.33e-03, grad_scale: 8.0 +2023-03-21 09:03:00,021 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106448.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:03:00,382 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.321e+02 1.800e+02 2.066e+02 2.441e+02 4.062e+02, threshold=4.133e+02, percent-clipped=0.0 +2023-03-21 09:03:03,085 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106454.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:03:03,474 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 09:03:03,550 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106455.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:03:04,757 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-03-21 09:03:07,045 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7117, 5.3034, 5.3586, 5.2960, 5.0855, 4.8377, 5.3967, 5.1257], + device='cuda:1'), covar=tensor([0.0483, 0.0361, 0.0361, 0.0502, 0.0334, 0.0390, 0.0314, 0.0484], + device='cuda:1'), in_proj_covar=tensor([0.0218, 0.0255, 0.0200, 0.0201, 0.0158, 0.0230, 0.0209, 0.0153], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:03:08,524 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 09:03:09,537 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 09:03:12,234 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1753, 2.6829, 3.2795, 3.2901, 2.8549, 2.7517, 3.2686, 2.5054], + device='cuda:1'), covar=tensor([0.0407, 0.0428, 0.0695, 0.0553, 0.0745, 0.0932, 0.0490, 0.2127], + device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0334, 0.0271, 0.0357, 0.0291, 0.0286, 0.0347, 0.0249], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:03:20,976 INFO [train.py:901] (1/2) Epoch 38, batch 2000, loss[loss=0.1386, simple_loss=0.2218, pruned_loss=0.02772, over 6670.00 frames. ], tot_loss[loss=0.1314, simple_loss=0.2127, pruned_loss=0.025, over 1443746.66 frames. ], batch size: 107, lr: 4.33e-03, grad_scale: 8.0 +2023-03-21 09:03:24,600 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=106496.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:03:26,038 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 09:03:28,087 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=106503.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:03:33,731 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6507, 4.1809, 4.2658, 4.3235, 4.3008, 4.1790, 4.4839, 3.9759], + device='cuda:1'), covar=tensor([0.0117, 0.0145, 0.0101, 0.0134, 0.0355, 0.0104, 0.0121, 0.0195], + device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0100, 0.0101, 0.0088, 0.0175, 0.0107, 0.0104, 0.0111], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:03:34,765 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106515.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:03:38,133 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 09:03:40,699 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8241, 4.0681, 3.8536, 4.1333, 3.6548, 4.0953, 4.3419, 4.3712], + device='cuda:1'), covar=tensor([0.0245, 0.0166, 0.0213, 0.0148, 0.0373, 0.0294, 0.0236, 0.0198], + device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0125, 0.0119, 0.0121, 0.0114, 0.0102, 0.0098, 0.0100], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:03:44,987 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 09:03:46,435 INFO [train.py:901] (1/2) Epoch 38, batch 2050, loss[loss=0.1425, simple_loss=0.2304, pruned_loss=0.02731, over 7284.00 frames. ], tot_loss[loss=0.1311, simple_loss=0.2123, pruned_loss=0.02497, over 1444470.99 frames. ], batch size: 66, lr: 4.33e-03, grad_scale: 8.0 +2023-03-21 09:03:46,586 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9101, 3.2149, 2.7274, 3.0882, 3.0077, 2.5925, 3.1631, 2.9640], + device='cuda:1'), covar=tensor([0.0812, 0.0445, 0.0989, 0.0969, 0.1140, 0.0777, 0.0704, 0.0964], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0058, 0.0066, 0.0059, 0.0056, 0.0061, 0.0055, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:03:49,101 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9527, 3.3693, 2.8159, 3.1795, 3.2613, 2.7761, 3.2877, 3.0432], + device='cuda:1'), covar=tensor([0.1183, 0.0546, 0.1032, 0.0927, 0.0706, 0.0885, 0.0595, 0.0987], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0058, 0.0066, 0.0059, 0.0056, 0.0061, 0.0055, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:03:51,412 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.246e+02 1.730e+02 1.940e+02 2.264e+02 3.831e+02, threshold=3.879e+02, percent-clipped=0.0 +2023-03-21 09:03:55,611 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2011, 2.2590, 2.4100, 2.0745, 2.0926, 2.2853, 2.0061, 1.7756], + device='cuda:1'), covar=tensor([0.0464, 0.0557, 0.0350, 0.0435, 0.0527, 0.0408, 0.0455, 0.0425], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0039, 0.0039, 0.0039, 0.0037, 0.0038, 0.0042, 0.0041], + device='cuda:1'), out_proj_covar=tensor([1.0121e-04, 1.0086e-04, 9.9512e-05, 9.8988e-05, 9.7158e-05, 9.7229e-05, + 1.0448e-04, 1.0493e-04], device='cuda:1') +2023-03-21 09:04:12,061 INFO [train.py:901] (1/2) Epoch 38, batch 2100, loss[loss=0.1277, simple_loss=0.2094, pruned_loss=0.02301, over 7275.00 frames. ], tot_loss[loss=0.1313, simple_loss=0.2123, pruned_loss=0.02519, over 1442958.45 frames. ], batch size: 77, lr: 4.33e-03, grad_scale: 8.0 +2023-03-21 09:04:20,469 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 09:04:22,916 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 09:04:30,497 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106624.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:04:37,934 INFO [train.py:901] (1/2) Epoch 38, batch 2150, loss[loss=0.1432, simple_loss=0.2277, pruned_loss=0.02933, over 7292.00 frames. ], tot_loss[loss=0.1316, simple_loss=0.2123, pruned_loss=0.02546, over 1440771.57 frames. ], batch size: 57, lr: 4.33e-03, grad_scale: 8.0 +2023-03-21 09:04:42,848 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.352e+02 1.839e+02 2.093e+02 2.461e+02 3.515e+02, threshold=4.185e+02, percent-clipped=0.0 +2023-03-21 09:05:04,247 INFO [train.py:901] (1/2) Epoch 38, batch 2200, loss[loss=0.1051, simple_loss=0.1817, pruned_loss=0.01421, over 6982.00 frames. ], tot_loss[loss=0.1319, simple_loss=0.2126, pruned_loss=0.02557, over 1439986.45 frames. ], batch size: 35, lr: 4.33e-03, grad_scale: 8.0 +2023-03-21 09:05:05,834 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106692.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:05:07,274 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 09:05:16,324 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6148, 2.9073, 2.6161, 2.7588, 2.8414, 2.4007, 2.8315, 2.7770], + device='cuda:1'), covar=tensor([0.0766, 0.0731, 0.0736, 0.1049, 0.1025, 0.0853, 0.0872, 0.0829], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0058, 0.0066, 0.0059, 0.0057, 0.0062, 0.0056, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:05:26,399 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106733.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:05:27,367 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2388, 2.5182, 2.5670, 2.3235, 2.2621, 2.3349, 2.2021, 1.8326], + device='cuda:1'), covar=tensor([0.0450, 0.0330, 0.0241, 0.0229, 0.0620, 0.0547, 0.0325, 0.0371], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0039, 0.0039, 0.0039, 0.0037, 0.0037, 0.0042, 0.0041], + device='cuda:1'), out_proj_covar=tensor([1.0031e-04, 1.0017e-04, 9.8853e-05, 9.8218e-05, 9.6300e-05, 9.6296e-05, + 1.0400e-04, 1.0429e-04], device='cuda:1') +2023-03-21 09:05:29,245 INFO [train.py:901] (1/2) Epoch 38, batch 2250, loss[loss=0.1207, simple_loss=0.2065, pruned_loss=0.01744, over 7351.00 frames. ], tot_loss[loss=0.1322, simple_loss=0.2129, pruned_loss=0.02568, over 1441575.51 frames. ], batch size: 51, lr: 4.33e-03, grad_scale: 8.0 +2023-03-21 09:05:34,241 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.029e+02 1.807e+02 2.184e+02 2.460e+02 5.015e+02, threshold=4.369e+02, percent-clipped=2.0 +2023-03-21 09:05:34,428 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3409, 2.2976, 2.4037, 3.5540, 1.8401, 3.5750, 1.4545, 3.4512], + device='cuda:1'), covar=tensor([0.0209, 0.1435, 0.1680, 0.0260, 0.3840, 0.0333, 0.1251, 0.0436], + device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0243, 0.0257, 0.0205, 0.0246, 0.0214, 0.0224, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:05:40,654 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1326, 2.5935, 3.1383, 2.8653, 3.0081, 2.8897, 2.6223, 2.9241], + device='cuda:1'), covar=tensor([0.0940, 0.0755, 0.1024, 0.1357, 0.0835, 0.0760, 0.1624, 0.1804], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0067, 0.0050, 0.0050, 0.0050, 0.0048, 0.0067, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 09:05:41,829 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-03-21 09:05:42,587 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 09:05:43,059 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 09:05:51,691 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=106781.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:05:55,683 INFO [train.py:901] (1/2) Epoch 38, batch 2300, loss[loss=0.1449, simple_loss=0.2251, pruned_loss=0.03232, over 7318.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.2127, pruned_loss=0.02547, over 1443544.47 frames. ], batch size: 59, lr: 4.33e-03, grad_scale: 8.0 +2023-03-21 09:05:56,686 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 09:06:06,578 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106810.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:06:21,173 INFO [train.py:901] (1/2) Epoch 38, batch 2350, loss[loss=0.1455, simple_loss=0.231, pruned_loss=0.03003, over 7299.00 frames. ], tot_loss[loss=0.1324, simple_loss=0.2135, pruned_loss=0.02572, over 1444956.57 frames. ], batch size: 80, lr: 4.33e-03, grad_scale: 8.0 +2023-03-21 09:06:26,822 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 9.316e+01 1.748e+02 2.140e+02 2.533e+02 7.476e+02, threshold=4.280e+02, percent-clipped=2.0 +2023-03-21 09:06:36,138 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9250, 3.0402, 2.7916, 3.9697, 1.9522, 4.0258, 1.7037, 3.4535], + device='cuda:1'), covar=tensor([0.0176, 0.1068, 0.1700, 0.0213, 0.4743, 0.0311, 0.1341, 0.0541], + device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0246, 0.0261, 0.0207, 0.0251, 0.0217, 0.0227, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:06:42,037 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 09:06:42,648 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7852, 3.8216, 2.9604, 3.4079, 2.8311, 2.1383, 1.7500, 3.8063], + device='cuda:1'), covar=tensor([0.0066, 0.0059, 0.0195, 0.0095, 0.0210, 0.0687, 0.0754, 0.0078], + device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0094, 0.0115, 0.0094, 0.0130, 0.0137, 0.0130, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 09:06:47,480 INFO [train.py:901] (1/2) Epoch 38, batch 2400, loss[loss=0.1365, simple_loss=0.2036, pruned_loss=0.03475, over 7251.00 frames. ], tot_loss[loss=0.1322, simple_loss=0.2133, pruned_loss=0.02559, over 1444656.22 frames. ], batch size: 45, lr: 4.32e-03, grad_scale: 8.0 +2023-03-21 09:06:47,665 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0680, 2.5428, 3.0770, 2.9740, 2.7344, 2.6482, 3.0844, 2.3005], + device='cuda:1'), covar=tensor([0.0614, 0.0451, 0.0663, 0.0666, 0.0665, 0.0919, 0.0630, 0.2288], + device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0332, 0.0269, 0.0355, 0.0290, 0.0284, 0.0345, 0.0248], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:06:48,501 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 09:06:51,598 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0107, 3.2371, 3.9754, 3.9418, 3.9901, 3.9220, 4.0702, 3.8468], + device='cuda:1'), covar=tensor([0.0025, 0.0119, 0.0032, 0.0031, 0.0031, 0.0033, 0.0037, 0.0048], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0069, 0.0057, 0.0056, 0.0055, 0.0060, 0.0048, 0.0077], + device='cuda:1'), out_proj_covar=tensor([8.2010e-05, 1.4100e-04, 1.0446e-04, 9.7184e-05, 9.4672e-05, 1.0564e-04, + 9.2102e-05, 1.4433e-04], device='cuda:1') +2023-03-21 09:06:58,539 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 09:06:59,617 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9532, 3.8240, 3.7548, 3.7186, 3.1732, 3.4748, 3.8279, 3.4353], + device='cuda:1'), covar=tensor([0.0239, 0.0241, 0.0183, 0.0271, 0.0894, 0.0236, 0.0347, 0.0295], + device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0100, 0.0101, 0.0088, 0.0175, 0.0106, 0.0103, 0.0111], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:07:01,617 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 09:07:05,195 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106924.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:07:13,218 INFO [train.py:901] (1/2) Epoch 38, batch 2450, loss[loss=0.1467, simple_loss=0.2396, pruned_loss=0.02686, over 7326.00 frames. ], tot_loss[loss=0.132, simple_loss=0.2133, pruned_loss=0.02537, over 1445056.83 frames. ], batch size: 59, lr: 4.32e-03, grad_scale: 8.0 +2023-03-21 09:07:18,719 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.280e+02 1.784e+02 2.053e+02 2.551e+02 4.549e+02, threshold=4.105e+02, percent-clipped=1.0 +2023-03-21 09:07:28,756 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 09:07:30,391 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=106972.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:07:31,412 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5373, 4.0396, 3.8878, 4.4700, 4.2465, 4.3964, 3.8130, 4.0084], + device='cuda:1'), covar=tensor([0.0857, 0.2542, 0.2607, 0.1154, 0.1048, 0.1366, 0.1034, 0.1389], + device='cuda:1'), in_proj_covar=tensor([0.0261, 0.0390, 0.0297, 0.0313, 0.0226, 0.0367, 0.0229, 0.0271], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:07:38,876 INFO [train.py:901] (1/2) Epoch 38, batch 2500, loss[loss=0.1267, simple_loss=0.2142, pruned_loss=0.0196, over 7325.00 frames. ], tot_loss[loss=0.1322, simple_loss=0.2137, pruned_loss=0.02536, over 1444473.55 frames. ], batch size: 61, lr: 4.32e-03, grad_scale: 8.0 +2023-03-21 09:07:40,485 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106992.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:07:48,214 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5480, 1.7844, 1.5921, 1.7009, 1.8288, 1.6684, 1.7024, 1.2869], + device='cuda:1'), covar=tensor([0.0143, 0.0136, 0.0284, 0.0129, 0.0141, 0.0126, 0.0179, 0.0199], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0036, 0.0035, 0.0036, 0.0035, 0.0034, 0.0037, 0.0045], + device='cuda:1'), out_proj_covar=tensor([4.2492e-05, 4.0044e-05, 3.9943e-05, 4.0154e-05, 3.8978e-05, 3.7611e-05, + 4.2056e-05, 4.9799e-05], device='cuda:1') +2023-03-21 09:07:54,690 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 09:08:05,291 INFO [train.py:901] (1/2) Epoch 38, batch 2550, loss[loss=0.1305, simple_loss=0.2158, pruned_loss=0.02262, over 7288.00 frames. ], tot_loss[loss=0.1327, simple_loss=0.214, pruned_loss=0.02566, over 1444774.90 frames. ], batch size: 66, lr: 4.32e-03, grad_scale: 8.0 +2023-03-21 09:08:05,855 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=107040.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:08:10,296 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 1.774e+02 2.109e+02 2.411e+02 4.197e+02, threshold=4.218e+02, percent-clipped=1.0 +2023-03-21 09:08:27,257 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3413, 4.2238, 3.7539, 3.8378, 3.3933, 2.4127, 2.0149, 4.3152], + device='cuda:1'), covar=tensor([0.0042, 0.0076, 0.0112, 0.0072, 0.0128, 0.0570, 0.0600, 0.0051], + device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0093, 0.0114, 0.0094, 0.0129, 0.0135, 0.0129, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 09:08:30,030 INFO [train.py:901] (1/2) Epoch 38, batch 2600, loss[loss=0.1323, simple_loss=0.2164, pruned_loss=0.02412, over 7320.00 frames. ], tot_loss[loss=0.1323, simple_loss=0.2136, pruned_loss=0.02552, over 1446581.83 frames. ], batch size: 59, lr: 4.32e-03, grad_scale: 8.0 +2023-03-21 09:08:39,046 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1671, 4.6657, 4.7160, 4.6526, 4.6704, 4.2628, 4.7263, 4.6065], + device='cuda:1'), covar=tensor([0.0529, 0.0402, 0.0353, 0.0505, 0.0301, 0.0416, 0.0336, 0.0432], + device='cuda:1'), in_proj_covar=tensor([0.0216, 0.0255, 0.0200, 0.0198, 0.0157, 0.0228, 0.0208, 0.0150], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:08:40,535 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107110.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:08:55,263 INFO [train.py:901] (1/2) Epoch 38, batch 2650, loss[loss=0.1461, simple_loss=0.231, pruned_loss=0.03058, over 6632.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.213, pruned_loss=0.02524, over 1444037.06 frames. ], batch size: 106, lr: 4.32e-03, grad_scale: 8.0 +2023-03-21 09:08:55,895 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107140.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 09:09:00,239 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 1.805e+02 2.120e+02 2.557e+02 4.024e+02, threshold=4.240e+02, percent-clipped=0.0 +2023-03-21 09:09:01,827 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107152.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:09:04,649 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=107158.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:09:20,369 INFO [train.py:901] (1/2) Epoch 38, batch 2700, loss[loss=0.1258, simple_loss=0.2093, pruned_loss=0.02113, over 7262.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.2129, pruned_loss=0.02537, over 1441268.15 frames. ], batch size: 64, lr: 4.32e-03, grad_scale: 8.0 +2023-03-21 09:09:26,649 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107201.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 09:09:31,750 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-03-21 09:09:32,471 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107213.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 09:09:44,595 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4744, 2.6322, 2.4834, 3.5107, 2.0152, 3.3916, 1.5216, 3.3292], + device='cuda:1'), covar=tensor([0.0191, 0.1247, 0.1772, 0.0219, 0.3563, 0.0310, 0.1254, 0.0444], + device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0246, 0.0260, 0.0206, 0.0249, 0.0216, 0.0227, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:09:44,937 INFO [train.py:901] (1/2) Epoch 38, batch 2750, loss[loss=0.1401, simple_loss=0.2202, pruned_loss=0.03, over 7336.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.2129, pruned_loss=0.02536, over 1441718.50 frames. ], batch size: 63, lr: 4.32e-03, grad_scale: 8.0 +2023-03-21 09:09:49,620 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 +2023-03-21 09:09:49,724 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.753e+02 2.094e+02 2.542e+02 3.915e+02, threshold=4.187e+02, percent-clipped=0.0 +2023-03-21 09:09:59,114 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9050, 3.8778, 3.0630, 3.4729, 2.8181, 2.2332, 1.9377, 3.8717], + device='cuda:1'), covar=tensor([0.0047, 0.0048, 0.0160, 0.0074, 0.0189, 0.0546, 0.0645, 0.0057], + device='cuda:1'), in_proj_covar=tensor([0.0112, 0.0092, 0.0113, 0.0093, 0.0128, 0.0134, 0.0129, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 09:10:03,200 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-03-21 09:10:09,086 INFO [train.py:901] (1/2) Epoch 38, batch 2800, loss[loss=0.1353, simple_loss=0.2146, pruned_loss=0.02794, over 7313.00 frames. ], tot_loss[loss=0.1322, simple_loss=0.2136, pruned_loss=0.02545, over 1442967.78 frames. ], batch size: 83, lr: 4.32e-03, grad_scale: 8.0 +2023-03-21 09:10:34,292 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 09:10:34,360 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 09:10:34,392 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 09:10:34,409 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 09:10:34,577 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 09:10:34,593 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 09:10:34,603 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 09:10:34,698 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 09:10:34,770 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 09:10:35,110 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 09:10:35,127 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 09:10:35,285 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 09:10:35,291 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 09:10:35,298 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 09:10:35,433 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 09:10:35,436 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 09:10:35,493 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 09:10:40,355 INFO [train.py:901] (1/2) Epoch 39, batch 0, loss[loss=0.1296, simple_loss=0.2075, pruned_loss=0.02578, over 7262.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.2075, pruned_loss=0.02578, over 7262.00 frames. ], batch size: 64, lr: 4.26e-03, grad_scale: 8.0 +2023-03-21 09:10:40,356 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 09:10:49,841 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9804, 3.0651, 2.3980, 3.6340, 2.3098, 2.8419, 1.6902, 2.5967], + device='cuda:1'), covar=tensor([0.0373, 0.0673, 0.2663, 0.0468, 0.0390, 0.0598, 0.4013, 0.1777], + device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0257, 0.0282, 0.0270, 0.0272, 0.0266, 0.0235, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:10:54,119 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7165, 5.0147, 5.0922, 5.0226, 4.7691, 4.5287, 5.0645, 4.7693], + device='cuda:1'), covar=tensor([0.0407, 0.0321, 0.0317, 0.0419, 0.0383, 0.0368, 0.0314, 0.0497], + device='cuda:1'), in_proj_covar=tensor([0.0217, 0.0257, 0.0201, 0.0200, 0.0159, 0.0229, 0.0209, 0.0151], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:11:04,174 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8462, 1.7048, 1.9761, 2.3689, 2.0239, 2.1575, 1.9111, 2.2657], + device='cuda:1'), covar=tensor([0.1406, 0.3050, 0.2631, 0.1143, 0.1002, 0.1228, 0.1297, 0.1626], + device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0077, 0.0070, 0.0062, 0.0061, 0.0060, 0.0101, 0.0066], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:11:05,880 INFO [train.py:935] (1/2) Epoch 39, validation: loss=0.1649, simple_loss=0.257, pruned_loss=0.03634, over 1622729.00 frames. +2023-03-21 09:11:05,881 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 09:11:12,876 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 09:11:22,873 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 09:11:23,859 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.279e+02 1.717e+02 1.963e+02 2.298e+02 3.449e+02, threshold=3.926e+02, percent-clipped=0.0 +2023-03-21 09:11:30,533 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 09:11:31,503 INFO [train.py:901] (1/2) Epoch 39, batch 50, loss[loss=0.1797, simple_loss=0.2553, pruned_loss=0.05208, over 6736.00 frames. ], tot_loss[loss=0.1314, simple_loss=0.212, pruned_loss=0.02539, over 323457.38 frames. ], batch size: 106, lr: 4.26e-03, grad_scale: 8.0 +2023-03-21 09:11:33,079 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 09:11:34,170 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6933, 1.5211, 1.7950, 2.1155, 1.8198, 2.0555, 1.4918, 2.0203], + device='cuda:1'), covar=tensor([0.1719, 0.3791, 0.1104, 0.1230, 0.1590, 0.1831, 0.1634, 0.1754], + device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0076, 0.0069, 0.0061, 0.0061, 0.0060, 0.0100, 0.0065], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:11:35,495 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 09:11:36,545 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5230, 4.0325, 3.8915, 4.4616, 4.2437, 4.3911, 3.9272, 4.0268], + device='cuda:1'), covar=tensor([0.0791, 0.2636, 0.2453, 0.1088, 0.0992, 0.1305, 0.0809, 0.1239], + device='cuda:1'), in_proj_covar=tensor([0.0259, 0.0391, 0.0295, 0.0311, 0.0227, 0.0364, 0.0227, 0.0270], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:11:52,910 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 09:11:53,358 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 09:11:53,500 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7433, 1.5882, 2.0247, 2.3415, 2.1132, 2.1479, 1.9352, 2.2757], + device='cuda:1'), covar=tensor([0.5317, 0.3573, 0.1499, 0.1712, 0.4247, 0.2438, 0.3146, 0.2195], + device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0076, 0.0069, 0.0061, 0.0060, 0.0060, 0.0100, 0.0065], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:11:57,311 INFO [train.py:901] (1/2) Epoch 39, batch 100, loss[loss=0.1498, simple_loss=0.2304, pruned_loss=0.03456, over 7325.00 frames. ], tot_loss[loss=0.1331, simple_loss=0.214, pruned_loss=0.02607, over 571564.05 frames. ], batch size: 75, lr: 4.26e-03, grad_scale: 8.0 +2023-03-21 09:12:15,948 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.043e+02 1.717e+02 1.993e+02 2.406e+02 7.451e+02, threshold=3.985e+02, percent-clipped=2.0 +2023-03-21 09:12:23,007 INFO [train.py:901] (1/2) Epoch 39, batch 150, loss[loss=0.1218, simple_loss=0.2045, pruned_loss=0.01958, over 7283.00 frames. ], tot_loss[loss=0.1313, simple_loss=0.2124, pruned_loss=0.02507, over 763884.09 frames. ], batch size: 77, lr: 4.26e-03, grad_scale: 8.0 +2023-03-21 09:12:40,178 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107496.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 09:12:46,180 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107508.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 09:12:47,215 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107510.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:12:48,615 INFO [train.py:901] (1/2) Epoch 39, batch 200, loss[loss=0.1328, simple_loss=0.2135, pruned_loss=0.02608, over 7316.00 frames. ], tot_loss[loss=0.1309, simple_loss=0.2126, pruned_loss=0.02454, over 914504.65 frames. ], batch size: 83, lr: 4.26e-03, grad_scale: 8.0 +2023-03-21 09:12:52,625 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 09:12:57,193 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 09:13:03,287 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 09:13:07,779 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.289e+02 1.722e+02 1.965e+02 2.423e+02 4.640e+02, threshold=3.930e+02, percent-clipped=2.0 +2023-03-21 09:13:14,178 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5801, 2.8626, 2.9643, 2.5457, 2.7119, 2.8379, 2.4209, 2.2721], + device='cuda:1'), covar=tensor([0.0446, 0.0401, 0.0189, 0.0338, 0.0546, 0.0338, 0.0247, 0.0282], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0038, 0.0038, 0.0038, 0.0035, 0.0036, 0.0041, 0.0040], + device='cuda:1'), out_proj_covar=tensor([9.8034e-05, 9.7820e-05, 9.7060e-05, 9.5569e-05, 9.3387e-05, 9.3052e-05, + 1.0119e-04, 1.0237e-04], device='cuda:1') +2023-03-21 09:13:14,709 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9744, 3.3655, 2.7562, 3.0866, 3.1739, 2.8903, 3.1106, 3.1482], + device='cuda:1'), covar=tensor([0.0795, 0.0577, 0.1061, 0.1423, 0.1136, 0.0712, 0.0890, 0.0791], + device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0059, 0.0067, 0.0060, 0.0057, 0.0062, 0.0056, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:13:15,075 INFO [train.py:901] (1/2) Epoch 39, batch 250, loss[loss=0.1282, simple_loss=0.2083, pruned_loss=0.02407, over 7341.00 frames. ], tot_loss[loss=0.1289, simple_loss=0.2106, pruned_loss=0.02363, over 1030501.07 frames. ], batch size: 49, lr: 4.25e-03, grad_scale: 8.0 +2023-03-21 09:13:16,559 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 09:13:19,136 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107571.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:13:19,638 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7187, 2.9239, 3.6267, 3.7167, 3.7230, 3.8216, 3.6070, 3.5849], + device='cuda:1'), covar=tensor([0.0028, 0.0125, 0.0031, 0.0029, 0.0031, 0.0024, 0.0056, 0.0049], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0068, 0.0056, 0.0055, 0.0054, 0.0059, 0.0048, 0.0076], + device='cuda:1'), out_proj_covar=tensor([8.1205e-05, 1.3835e-04, 1.0179e-04, 9.5067e-05, 9.2129e-05, 1.0363e-04, + 9.1320e-05, 1.4200e-04], device='cuda:1') +2023-03-21 09:13:37,329 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 09:13:40,234 INFO [train.py:901] (1/2) Epoch 39, batch 300, loss[loss=0.1205, simple_loss=0.2113, pruned_loss=0.01482, over 7295.00 frames. ], tot_loss[loss=0.1291, simple_loss=0.2109, pruned_loss=0.02367, over 1122284.42 frames. ], batch size: 66, lr: 4.25e-03, grad_scale: 8.0 +2023-03-21 09:13:45,480 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4194, 3.6305, 3.4598, 3.5561, 3.4312, 3.1321, 3.5135, 3.6415], + device='cuda:1'), covar=tensor([0.0444, 0.0256, 0.0356, 0.0335, 0.0454, 0.0614, 0.0507, 0.0410], + device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0123, 0.0117, 0.0121, 0.0111, 0.0101, 0.0096, 0.0098], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:13:46,373 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 09:13:59,555 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.007e+02 1.763e+02 1.989e+02 2.300e+02 3.698e+02, threshold=3.979e+02, percent-clipped=0.0 +2023-03-21 09:14:01,640 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3840, 4.9707, 5.0050, 4.9749, 4.8099, 4.4792, 5.0030, 4.8178], + device='cuda:1'), covar=tensor([0.0531, 0.0361, 0.0404, 0.0483, 0.0368, 0.0437, 0.0361, 0.0479], + device='cuda:1'), in_proj_covar=tensor([0.0217, 0.0256, 0.0201, 0.0202, 0.0160, 0.0229, 0.0209, 0.0150], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:14:04,727 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107659.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:14:06,587 INFO [train.py:901] (1/2) Epoch 39, batch 350, loss[loss=0.1322, simple_loss=0.2171, pruned_loss=0.02362, over 7345.00 frames. ], tot_loss[loss=0.1289, simple_loss=0.2104, pruned_loss=0.02372, over 1194152.33 frames. ], batch size: 75, lr: 4.25e-03, grad_scale: 8.0 +2023-03-21 09:14:13,109 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5290, 2.7942, 2.7721, 2.3784, 2.5684, 2.5919, 2.1717, 2.0028], + device='cuda:1'), covar=tensor([0.0374, 0.0427, 0.0275, 0.0277, 0.0693, 0.0450, 0.0571, 0.0419], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0038, 0.0038, 0.0038, 0.0036, 0.0036, 0.0041, 0.0041], + device='cuda:1'), out_proj_covar=tensor([9.9087e-05, 9.8542e-05, 9.7356e-05, 9.5992e-05, 9.3973e-05, 9.3609e-05, + 1.0220e-04, 1.0301e-04], device='cuda:1') +2023-03-21 09:14:22,319 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 09:14:31,993 INFO [train.py:901] (1/2) Epoch 39, batch 400, loss[loss=0.1198, simple_loss=0.1965, pruned_loss=0.02148, over 7227.00 frames. ], tot_loss[loss=0.1291, simple_loss=0.2102, pruned_loss=0.02406, over 1248352.63 frames. ], batch size: 45, lr: 4.25e-03, grad_scale: 16.0 +2023-03-21 09:14:35,583 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107720.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:14:43,795 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6193, 2.4808, 2.5670, 3.7569, 2.0846, 3.5373, 1.4891, 3.3293], + device='cuda:1'), covar=tensor([0.0197, 0.1490, 0.1948, 0.0279, 0.3799, 0.0301, 0.1353, 0.0393], + device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0248, 0.0264, 0.0209, 0.0252, 0.0219, 0.0229, 0.0229], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:14:44,772 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107737.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:14:50,589 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.205e+02 1.824e+02 2.104e+02 2.384e+02 4.420e+02, threshold=4.207e+02, percent-clipped=2.0 +2023-03-21 09:14:57,691 INFO [train.py:901] (1/2) Epoch 39, batch 450, loss[loss=0.128, simple_loss=0.2126, pruned_loss=0.02175, over 7307.00 frames. ], tot_loss[loss=0.1299, simple_loss=0.2111, pruned_loss=0.02438, over 1291015.45 frames. ], batch size: 68, lr: 4.25e-03, grad_scale: 16.0 +2023-03-21 09:14:59,751 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5221, 4.0999, 4.0878, 4.2431, 4.0921, 4.0366, 4.3867, 3.8562], + device='cuda:1'), covar=tensor([0.0159, 0.0158, 0.0124, 0.0137, 0.0468, 0.0124, 0.0145, 0.0187], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0101, 0.0103, 0.0089, 0.0179, 0.0108, 0.0105, 0.0113], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:15:03,160 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 09:15:03,173 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 09:15:14,951 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107796.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 09:15:15,989 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107798.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 09:15:21,321 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107808.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:15:23,748 INFO [train.py:901] (1/2) Epoch 39, batch 500, loss[loss=0.1182, simple_loss=0.1893, pruned_loss=0.02357, over 6971.00 frames. ], tot_loss[loss=0.1303, simple_loss=0.2113, pruned_loss=0.02468, over 1324141.04 frames. ], batch size: 35, lr: 4.25e-03, grad_scale: 16.0 +2023-03-21 09:15:37,396 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 09:15:38,897 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 09:15:39,383 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 09:15:39,931 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=107844.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 09:15:41,917 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 09:15:42,403 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.285e+02 1.745e+02 2.038e+02 2.318e+02 3.534e+02, threshold=4.076e+02, percent-clipped=0.0 +2023-03-21 09:15:45,891 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=107856.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:15:46,352 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 09:15:49,358 INFO [train.py:901] (1/2) Epoch 39, batch 550, loss[loss=0.1331, simple_loss=0.2139, pruned_loss=0.02615, over 7351.00 frames. ], tot_loss[loss=0.131, simple_loss=0.2122, pruned_loss=0.02484, over 1351473.57 frames. ], batch size: 44, lr: 4.25e-03, grad_scale: 8.0 +2023-03-21 09:15:50,905 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107866.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:15:57,751 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 09:16:04,055 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8542, 3.2464, 2.7865, 3.1244, 3.1334, 2.8698, 3.1558, 2.8461], + device='cuda:1'), covar=tensor([0.0588, 0.0628, 0.0943, 0.0892, 0.0918, 0.0631, 0.0779, 0.1478], + device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0058, 0.0066, 0.0059, 0.0056, 0.0062, 0.0055, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:16:06,953 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 09:16:10,386 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 +2023-03-21 09:16:10,483 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 09:16:15,525 INFO [train.py:901] (1/2) Epoch 39, batch 600, loss[loss=0.1287, simple_loss=0.2182, pruned_loss=0.01965, over 7245.00 frames. ], tot_loss[loss=0.1313, simple_loss=0.2125, pruned_loss=0.02503, over 1373349.73 frames. ], batch size: 89, lr: 4.25e-03, grad_scale: 8.0 +2023-03-21 09:16:17,507 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 09:16:33,532 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 09:16:33,983 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.697e+02 2.000e+02 2.452e+02 4.115e+02, threshold=4.000e+02, percent-clipped=1.0 +2023-03-21 09:16:40,503 INFO [train.py:901] (1/2) Epoch 39, batch 650, loss[loss=0.1329, simple_loss=0.2138, pruned_loss=0.02598, over 7308.00 frames. ], tot_loss[loss=0.131, simple_loss=0.2125, pruned_loss=0.02478, over 1390057.44 frames. ], batch size: 80, lr: 4.25e-03, grad_scale: 8.0 +2023-03-21 09:16:41,977 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 09:16:50,281 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1198, 2.7807, 3.1151, 3.0381, 2.8753, 2.7379, 3.0532, 2.3919], + device='cuda:1'), covar=tensor([0.0420, 0.0475, 0.0657, 0.0542, 0.0686, 0.0902, 0.0562, 0.2138], + device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0333, 0.0270, 0.0353, 0.0288, 0.0283, 0.0347, 0.0246], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:16:59,571 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 09:17:10,689 INFO [train.py:901] (1/2) Epoch 39, batch 700, loss[loss=0.1388, simple_loss=0.2215, pruned_loss=0.02803, over 7324.00 frames. ], tot_loss[loss=0.1311, simple_loss=0.2127, pruned_loss=0.02476, over 1403605.52 frames. ], batch size: 83, lr: 4.25e-03, grad_scale: 8.0 +2023-03-21 09:17:11,756 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108015.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:17:13,229 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 09:17:29,088 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.255e+02 1.760e+02 2.052e+02 2.429e+02 4.118e+02, threshold=4.103e+02, percent-clipped=1.0 +2023-03-21 09:17:34,862 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5287, 2.9851, 2.5243, 2.8274, 2.9596, 2.5690, 2.9070, 2.6152], + device='cuda:1'), covar=tensor([0.0896, 0.0493, 0.1053, 0.1039, 0.1140, 0.0829, 0.0763, 0.1274], + device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0058, 0.0066, 0.0059, 0.0057, 0.0062, 0.0056, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:17:36,197 INFO [train.py:901] (1/2) Epoch 39, batch 750, loss[loss=0.1265, simple_loss=0.2132, pruned_loss=0.01992, over 7137.00 frames. ], tot_loss[loss=0.1305, simple_loss=0.2118, pruned_loss=0.02463, over 1410949.59 frames. ], batch size: 98, lr: 4.25e-03, grad_scale: 8.0 +2023-03-21 09:17:38,245 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 09:17:38,766 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 09:17:51,864 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108093.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 09:17:52,739 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 09:17:56,694 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 09:18:01,649 INFO [train.py:901] (1/2) Epoch 39, batch 800, loss[loss=0.1759, simple_loss=0.2483, pruned_loss=0.05176, over 6755.00 frames. ], tot_loss[loss=0.1312, simple_loss=0.2124, pruned_loss=0.02504, over 1418936.08 frames. ], batch size: 107, lr: 4.24e-03, grad_scale: 8.0 +2023-03-21 09:18:02,664 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 09:18:03,086 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-03-21 09:18:03,665 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 09:18:03,814 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9936, 3.4302, 2.9442, 3.2627, 3.3585, 2.9810, 3.2039, 3.2513], + device='cuda:1'), covar=tensor([0.0520, 0.0536, 0.0885, 0.0874, 0.0973, 0.0831, 0.0712, 0.0722], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0058, 0.0065, 0.0058, 0.0056, 0.0061, 0.0055, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:18:13,784 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 09:18:20,773 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.260e+02 1.831e+02 2.034e+02 2.384e+02 3.789e+02, threshold=4.068e+02, percent-clipped=0.0 +2023-03-21 09:18:22,454 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5510, 2.7655, 2.4624, 2.7418, 2.8589, 2.5391, 2.6833, 2.6463], + device='cuda:1'), covar=tensor([0.0855, 0.0668, 0.1046, 0.0911, 0.0577, 0.0608, 0.1115, 0.0905], + device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0058, 0.0065, 0.0058, 0.0056, 0.0061, 0.0055, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:18:27,267 INFO [train.py:901] (1/2) Epoch 39, batch 850, loss[loss=0.1413, simple_loss=0.219, pruned_loss=0.03178, over 7291.00 frames. ], tot_loss[loss=0.1308, simple_loss=0.2123, pruned_loss=0.02465, over 1425667.15 frames. ], batch size: 77, lr: 4.24e-03, grad_scale: 8.0 +2023-03-21 09:18:29,473 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108166.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:18:33,932 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 09:18:34,408 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 09:18:39,418 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 09:18:43,621 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 09:18:53,335 INFO [train.py:901] (1/2) Epoch 39, batch 900, loss[loss=0.1361, simple_loss=0.2145, pruned_loss=0.02885, over 7257.00 frames. ], tot_loss[loss=0.13, simple_loss=0.2114, pruned_loss=0.02428, over 1427971.88 frames. ], batch size: 64, lr: 4.24e-03, grad_scale: 8.0 +2023-03-21 09:18:53,861 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=108214.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:18:56,419 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6524, 3.9649, 3.7072, 3.9554, 3.5077, 3.9283, 4.2834, 4.2708], + device='cuda:1'), covar=tensor([0.0267, 0.0158, 0.0262, 0.0168, 0.0454, 0.0355, 0.0186, 0.0174], + device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0125, 0.0119, 0.0123, 0.0114, 0.0102, 0.0097, 0.0100], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:19:12,501 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.190e+02 1.645e+02 1.970e+02 2.398e+02 5.896e+02, threshold=3.941e+02, percent-clipped=2.0 +2023-03-21 09:19:19,541 INFO [train.py:901] (1/2) Epoch 39, batch 950, loss[loss=0.1378, simple_loss=0.2206, pruned_loss=0.02752, over 7241.00 frames. ], tot_loss[loss=0.1305, simple_loss=0.2117, pruned_loss=0.02467, over 1429833.87 frames. ], batch size: 55, lr: 4.24e-03, grad_scale: 8.0 +2023-03-21 09:19:20,055 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 09:19:25,760 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7247, 3.8313, 3.1614, 4.1281, 3.5171, 3.8042, 2.1161, 3.1872], + device='cuda:1'), covar=tensor([0.0582, 0.0676, 0.1841, 0.0467, 0.0449, 0.0493, 0.2977, 0.1615], + device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0257, 0.0283, 0.0271, 0.0270, 0.0264, 0.0236, 0.0259], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:19:43,321 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 09:19:44,854 INFO [train.py:901] (1/2) Epoch 39, batch 1000, loss[loss=0.1383, simple_loss=0.2212, pruned_loss=0.02766, over 7255.00 frames. ], tot_loss[loss=0.1301, simple_loss=0.2108, pruned_loss=0.02464, over 1429494.87 frames. ], batch size: 55, lr: 4.24e-03, grad_scale: 8.0 +2023-03-21 09:19:45,912 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108315.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:19:51,491 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108325.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:20:03,595 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0205, 2.8118, 3.2991, 3.1309, 3.1849, 3.1443, 2.7659, 3.1754], + device='cuda:1'), covar=tensor([0.1641, 0.0809, 0.0839, 0.1192, 0.0883, 0.0959, 0.2077, 0.1137], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0068, 0.0051, 0.0051, 0.0051, 0.0049, 0.0068, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 09:20:04,450 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.251e+02 1.792e+02 2.080e+02 2.512e+02 3.266e+02, threshold=4.160e+02, percent-clipped=0.0 +2023-03-21 09:20:04,986 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 09:20:06,315 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 +2023-03-21 09:20:09,808 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 +2023-03-21 09:20:10,960 INFO [train.py:901] (1/2) Epoch 39, batch 1050, loss[loss=0.1155, simple_loss=0.2027, pruned_loss=0.01412, over 7222.00 frames. ], tot_loss[loss=0.1304, simple_loss=0.211, pruned_loss=0.02484, over 1432622.51 frames. ], batch size: 50, lr: 4.24e-03, grad_scale: 8.0 +2023-03-21 09:20:11,044 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=108363.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:20:17,303 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-03-21 09:20:18,652 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9475, 2.7283, 2.6450, 4.0018, 2.0342, 3.7524, 1.5758, 3.4217], + device='cuda:1'), covar=tensor([0.0224, 0.1369, 0.1760, 0.0217, 0.3787, 0.0307, 0.1391, 0.0479], + device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0245, 0.0261, 0.0206, 0.0249, 0.0217, 0.0227, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:20:22,700 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108386.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:20:26,115 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108393.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:20:26,949 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 09:20:31,771 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 09:20:36,757 INFO [train.py:901] (1/2) Epoch 39, batch 1100, loss[loss=0.1471, simple_loss=0.2356, pruned_loss=0.02928, over 7288.00 frames. ], tot_loss[loss=0.1303, simple_loss=0.2111, pruned_loss=0.02479, over 1435089.38 frames. ], batch size: 68, lr: 4.24e-03, grad_scale: 8.0 +2023-03-21 09:20:43,205 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.15 vs. limit=5.0 +2023-03-21 09:20:47,012 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108432.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:20:51,398 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=108441.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:20:55,529 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7001, 2.3708, 2.8136, 2.6794, 2.7208, 2.6873, 2.3998, 2.9134], + device='cuda:1'), covar=tensor([0.1648, 0.0894, 0.1205, 0.1634, 0.1106, 0.0945, 0.1869, 0.1255], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0068, 0.0051, 0.0051, 0.0051, 0.0049, 0.0068, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 09:20:55,869 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.198e+02 1.659e+02 1.991e+02 2.465e+02 4.884e+02, threshold=3.983e+02, percent-clipped=1.0 +2023-03-21 09:21:00,959 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 09:21:01,480 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 09:21:02,524 INFO [train.py:901] (1/2) Epoch 39, batch 1150, loss[loss=0.1129, simple_loss=0.1923, pruned_loss=0.01679, over 7140.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.2108, pruned_loss=0.02425, over 1438662.92 frames. ], batch size: 41, lr: 4.24e-03, grad_scale: 8.0 +2023-03-21 09:21:08,234 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3571, 3.4873, 2.3704, 3.8981, 3.0469, 3.3789, 1.6763, 2.4868], + device='cuda:1'), covar=tensor([0.0586, 0.1029, 0.2791, 0.0580, 0.0520, 0.0600, 0.3818, 0.1784], + device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0258, 0.0283, 0.0271, 0.0271, 0.0266, 0.0235, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:21:10,312 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2461, 2.9340, 3.2415, 3.1876, 2.9291, 2.8754, 3.3386, 2.4025], + device='cuda:1'), covar=tensor([0.0512, 0.0596, 0.0600, 0.0754, 0.0690, 0.0937, 0.0806, 0.2296], + device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0333, 0.0270, 0.0353, 0.0287, 0.0284, 0.0345, 0.0246], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:21:15,061 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 09:21:15,560 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 09:21:18,358 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108493.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 09:21:28,201 INFO [train.py:901] (1/2) Epoch 39, batch 1200, loss[loss=0.1306, simple_loss=0.2151, pruned_loss=0.02308, over 7302.00 frames. ], tot_loss[loss=0.1297, simple_loss=0.2107, pruned_loss=0.02433, over 1439747.77 frames. ], batch size: 86, lr: 4.24e-03, grad_scale: 8.0 +2023-03-21 09:21:35,502 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7831, 2.9214, 3.7341, 3.7632, 3.7558, 3.8481, 3.7313, 3.6564], + device='cuda:1'), covar=tensor([0.0032, 0.0144, 0.0036, 0.0033, 0.0036, 0.0031, 0.0043, 0.0054], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0069, 0.0057, 0.0056, 0.0055, 0.0061, 0.0049, 0.0078], + device='cuda:1'), out_proj_covar=tensor([8.2175e-05, 1.4024e-04, 1.0425e-04, 9.7723e-05, 9.3835e-05, 1.0752e-04, + 9.2992e-05, 1.4608e-04], device='cuda:1') +2023-03-21 09:21:47,344 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.212e+02 1.679e+02 2.002e+02 2.337e+02 3.953e+02, threshold=4.003e+02, percent-clipped=0.0 +2023-03-21 09:21:47,883 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 09:21:53,858 INFO [train.py:901] (1/2) Epoch 39, batch 1250, loss[loss=0.1357, simple_loss=0.2148, pruned_loss=0.02828, over 7375.00 frames. ], tot_loss[loss=0.1305, simple_loss=0.2114, pruned_loss=0.02477, over 1440824.76 frames. ], batch size: 65, lr: 4.24e-03, grad_scale: 8.0 +2023-03-21 09:22:01,063 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108577.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:22:10,976 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 09:22:14,912 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 09:22:15,957 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 09:22:20,633 INFO [train.py:901] (1/2) Epoch 39, batch 1300, loss[loss=0.1278, simple_loss=0.2058, pruned_loss=0.02488, over 7231.00 frames. ], tot_loss[loss=0.1303, simple_loss=0.2114, pruned_loss=0.02455, over 1443198.37 frames. ], batch size: 45, lr: 4.23e-03, grad_scale: 8.0 +2023-03-21 09:22:29,541 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 +2023-03-21 09:22:33,458 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108638.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:22:39,206 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.296e+02 1.751e+02 2.074e+02 2.340e+02 4.099e+02, threshold=4.148e+02, percent-clipped=1.0 +2023-03-21 09:22:40,246 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 09:22:42,276 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 09:22:46,464 INFO [train.py:901] (1/2) Epoch 39, batch 1350, loss[loss=0.1404, simple_loss=0.2218, pruned_loss=0.02953, over 7258.00 frames. ], tot_loss[loss=0.1301, simple_loss=0.2112, pruned_loss=0.02452, over 1443415.25 frames. ], batch size: 55, lr: 4.23e-03, grad_scale: 8.0 +2023-03-21 09:22:47,003 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 09:22:55,612 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108681.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:22:56,583 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 09:23:10,912 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 +2023-03-21 09:23:11,794 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5881, 1.9055, 1.5882, 1.7069, 1.9267, 1.7353, 1.7290, 1.2269], + device='cuda:1'), covar=tensor([0.0262, 0.0178, 0.0309, 0.0171, 0.0118, 0.0222, 0.0173, 0.0215], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0036, 0.0035, 0.0036, 0.0035, 0.0033, 0.0037, 0.0044], + device='cuda:1'), out_proj_covar=tensor([4.1823e-05, 4.0221e-05, 3.9187e-05, 4.0057e-05, 3.8656e-05, 3.6995e-05, + 4.1587e-05, 4.9303e-05], device='cuda:1') +2023-03-21 09:23:12,157 INFO [train.py:901] (1/2) Epoch 39, batch 1400, loss[loss=0.1351, simple_loss=0.2159, pruned_loss=0.02718, over 7333.00 frames. ], tot_loss[loss=0.1298, simple_loss=0.211, pruned_loss=0.02425, over 1443608.77 frames. ], batch size: 59, lr: 4.23e-03, grad_scale: 8.0 +2023-03-21 09:23:12,349 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1646, 2.8326, 3.3496, 3.2611, 2.9977, 2.9032, 3.4926, 2.5640], + device='cuda:1'), covar=tensor([0.0400, 0.0462, 0.0716, 0.0617, 0.0640, 0.0982, 0.0616, 0.2088], + device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0336, 0.0272, 0.0356, 0.0289, 0.0287, 0.0347, 0.0248], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:23:19,471 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5910, 2.3190, 2.7751, 2.6617, 2.7552, 2.7331, 2.3289, 2.7934], + device='cuda:1'), covar=tensor([0.1706, 0.0772, 0.0811, 0.1097, 0.0654, 0.0650, 0.1841, 0.1228], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0066, 0.0050, 0.0050, 0.0049, 0.0047, 0.0066, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 09:23:27,741 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 09:23:31,340 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.707e+02 2.003e+02 2.347e+02 5.252e+02, threshold=4.006e+02, percent-clipped=1.0 +2023-03-21 09:23:37,761 INFO [train.py:901] (1/2) Epoch 39, batch 1450, loss[loss=0.122, simple_loss=0.2104, pruned_loss=0.01676, over 7265.00 frames. ], tot_loss[loss=0.1304, simple_loss=0.2119, pruned_loss=0.02444, over 1443846.26 frames. ], batch size: 64, lr: 4.23e-03, grad_scale: 8.0 +2023-03-21 09:23:51,099 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108788.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 09:23:52,633 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108791.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:23:52,997 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 09:24:03,877 INFO [train.py:901] (1/2) Epoch 39, batch 1500, loss[loss=0.09993, simple_loss=0.1727, pruned_loss=0.01357, over 7028.00 frames. ], tot_loss[loss=0.1302, simple_loss=0.2115, pruned_loss=0.02441, over 1444452.00 frames. ], batch size: 35, lr: 4.23e-03, grad_scale: 8.0 +2023-03-21 09:24:09,001 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4601, 1.7694, 1.4990, 1.6374, 1.7305, 1.6368, 1.5887, 1.1994], + device='cuda:1'), covar=tensor([0.0239, 0.0163, 0.0248, 0.0166, 0.0173, 0.0137, 0.0134, 0.0221], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0036, 0.0034, 0.0036, 0.0035, 0.0033, 0.0037, 0.0044], + device='cuda:1'), out_proj_covar=tensor([4.1671e-05, 4.0138e-05, 3.9064e-05, 4.0157e-05, 3.8497e-05, 3.6857e-05, + 4.1454e-05, 4.9228e-05], device='cuda:1') +2023-03-21 09:24:09,373 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 09:24:22,946 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.726e+02 1.983e+02 2.350e+02 3.872e+02, threshold=3.966e+02, percent-clipped=0.0 +2023-03-21 09:24:24,095 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108852.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:24:29,435 INFO [train.py:901] (1/2) Epoch 39, batch 1550, loss[loss=0.1034, simple_loss=0.1771, pruned_loss=0.01489, over 7046.00 frames. ], tot_loss[loss=0.1304, simple_loss=0.2117, pruned_loss=0.02453, over 1444244.03 frames. ], batch size: 35, lr: 4.23e-03, grad_scale: 8.0 +2023-03-21 09:24:33,179 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0239, 3.0686, 2.2483, 3.4669, 2.4357, 2.8905, 1.5225, 2.2317], + device='cuda:1'), covar=tensor([0.0509, 0.0750, 0.2991, 0.0791, 0.0450, 0.0784, 0.4116, 0.1929], + device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0257, 0.0280, 0.0271, 0.0271, 0.0265, 0.0234, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:24:34,656 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 09:24:55,264 INFO [train.py:901] (1/2) Epoch 39, batch 1600, loss[loss=0.1016, simple_loss=0.1837, pruned_loss=0.009787, over 6969.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.211, pruned_loss=0.0241, over 1442877.79 frames. ], batch size: 35, lr: 4.23e-03, grad_scale: 8.0 +2023-03-21 09:25:01,593 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4042, 4.1799, 3.5918, 3.9340, 3.4527, 2.2900, 1.9886, 4.3688], + device='cuda:1'), covar=tensor([0.0046, 0.0071, 0.0132, 0.0079, 0.0135, 0.0645, 0.0636, 0.0050], + device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0094, 0.0113, 0.0094, 0.0129, 0.0136, 0.0129, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 09:25:04,880 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 09:25:05,419 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 09:25:05,987 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108933.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:25:08,397 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 09:25:14,451 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 1.678e+02 2.010e+02 2.437e+02 3.858e+02, threshold=4.020e+02, percent-clipped=0.0 +2023-03-21 09:25:18,601 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 09:25:20,940 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-03-21 09:25:21,626 INFO [train.py:901] (1/2) Epoch 39, batch 1650, loss[loss=0.1321, simple_loss=0.2126, pruned_loss=0.02583, over 7248.00 frames. ], tot_loss[loss=0.13, simple_loss=0.2115, pruned_loss=0.02429, over 1443839.28 frames. ], batch size: 47, lr: 4.23e-03, grad_scale: 8.0 +2023-03-21 09:25:22,653 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 09:25:30,127 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 09:25:30,699 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108981.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:25:47,279 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 09:25:47,792 INFO [train.py:901] (1/2) Epoch 39, batch 1700, loss[loss=0.1242, simple_loss=0.2061, pruned_loss=0.02117, over 7294.00 frames. ], tot_loss[loss=0.1301, simple_loss=0.2115, pruned_loss=0.02433, over 1442115.54 frames. ], batch size: 68, lr: 4.23e-03, grad_scale: 8.0 +2023-03-21 09:25:51,859 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 09:25:56,092 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=109029.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:26:00,141 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109037.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 09:26:03,633 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 09:26:05,013 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-03-21 09:26:07,084 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.192e+02 1.706e+02 2.028e+02 2.336e+02 4.447e+02, threshold=4.056e+02, percent-clipped=2.0 +2023-03-21 09:26:13,718 INFO [train.py:901] (1/2) Epoch 39, batch 1750, loss[loss=0.1385, simple_loss=0.2226, pruned_loss=0.02725, over 7335.00 frames. ], tot_loss[loss=0.1303, simple_loss=0.2118, pruned_loss=0.02439, over 1443129.86 frames. ], batch size: 54, lr: 4.23e-03, grad_scale: 8.0 +2023-03-21 09:26:26,392 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109088.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:26:27,760 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 09:26:28,273 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 09:26:28,625 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 +2023-03-21 09:26:32,031 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109098.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 09:26:39,418 INFO [train.py:901] (1/2) Epoch 39, batch 1800, loss[loss=0.1166, simple_loss=0.2012, pruned_loss=0.01601, over 7347.00 frames. ], tot_loss[loss=0.1302, simple_loss=0.2115, pruned_loss=0.02443, over 1442023.93 frames. ], batch size: 44, lr: 4.22e-03, grad_scale: 8.0 +2023-03-21 09:26:49,036 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 +2023-03-21 09:26:51,052 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 09:26:51,639 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=109136.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:26:57,077 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109147.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:26:58,454 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.276e+02 1.788e+02 2.056e+02 2.387e+02 1.604e+03, threshold=4.112e+02, percent-clipped=1.0 +2023-03-21 09:27:04,193 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 09:27:05,201 INFO [train.py:901] (1/2) Epoch 39, batch 1850, loss[loss=0.1202, simple_loss=0.2044, pruned_loss=0.01804, over 7311.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.2109, pruned_loss=0.02412, over 1442288.09 frames. ], batch size: 86, lr: 4.22e-03, grad_scale: 8.0 +2023-03-21 09:27:14,719 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 09:27:21,504 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109194.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:27:31,328 INFO [train.py:901] (1/2) Epoch 39, batch 1900, loss[loss=0.1527, simple_loss=0.2288, pruned_loss=0.03833, over 7359.00 frames. ], tot_loss[loss=0.1298, simple_loss=0.211, pruned_loss=0.02425, over 1443041.97 frames. ], batch size: 54, lr: 4.22e-03, grad_scale: 8.0 +2023-03-21 09:27:32,349 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 09:27:42,179 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109233.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:27:50,589 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.101e+02 1.796e+02 1.987e+02 2.391e+02 7.670e+02, threshold=3.974e+02, percent-clipped=1.0 +2023-03-21 09:27:53,369 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109255.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:27:54,086 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.50 vs. limit=5.0 +2023-03-21 09:27:57,245 INFO [train.py:901] (1/2) Epoch 39, batch 1950, loss[loss=0.1383, simple_loss=0.2101, pruned_loss=0.03329, over 7291.00 frames. ], tot_loss[loss=0.1297, simple_loss=0.2107, pruned_loss=0.02436, over 1440876.94 frames. ], batch size: 49, lr: 4.22e-03, grad_scale: 8.0 +2023-03-21 09:27:57,756 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 09:28:06,991 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=109281.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:28:08,939 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 09:28:13,367 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 09:28:13,517 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109294.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:28:13,892 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 09:28:23,528 INFO [train.py:901] (1/2) Epoch 39, batch 2000, loss[loss=0.1208, simple_loss=0.2024, pruned_loss=0.01955, over 7308.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.2109, pruned_loss=0.02413, over 1442612.31 frames. ], batch size: 49, lr: 4.22e-03, grad_scale: 8.0 +2023-03-21 09:28:26,173 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109318.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:28:30,047 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 09:28:40,069 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 09:28:42,038 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.673e+02 2.043e+02 2.333e+02 6.605e+02, threshold=4.086e+02, percent-clipped=1.0 +2023-03-21 09:28:45,393 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109355.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:28:48,755 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 09:28:49,199 INFO [train.py:901] (1/2) Epoch 39, batch 2050, loss[loss=0.1325, simple_loss=0.2142, pruned_loss=0.02541, over 7227.00 frames. ], tot_loss[loss=0.1307, simple_loss=0.2122, pruned_loss=0.02462, over 1443237.31 frames. ], batch size: 93, lr: 4.22e-03, grad_scale: 8.0 +2023-03-21 09:28:57,396 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109379.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 09:29:04,842 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109393.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 09:29:15,137 INFO [train.py:901] (1/2) Epoch 39, batch 2100, loss[loss=0.1267, simple_loss=0.2112, pruned_loss=0.02107, over 7273.00 frames. ], tot_loss[loss=0.1303, simple_loss=0.2117, pruned_loss=0.02449, over 1442305.34 frames. ], batch size: 77, lr: 4.22e-03, grad_scale: 8.0 +2023-03-21 09:29:22,176 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 09:29:24,041 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-03-21 09:29:25,608 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 09:29:32,952 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109447.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:29:34,358 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.266e+02 1.657e+02 1.969e+02 2.192e+02 3.456e+02, threshold=3.938e+02, percent-clipped=0.0 +2023-03-21 09:29:41,177 INFO [train.py:901] (1/2) Epoch 39, batch 2150, loss[loss=0.1253, simple_loss=0.2096, pruned_loss=0.02046, over 7287.00 frames. ], tot_loss[loss=0.1307, simple_loss=0.2121, pruned_loss=0.02462, over 1444424.81 frames. ], batch size: 77, lr: 4.22e-03, grad_scale: 8.0 +2023-03-21 09:29:45,985 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4860, 3.9946, 3.9157, 4.4433, 4.3293, 4.3557, 3.9444, 3.9493], + device='cuda:1'), covar=tensor([0.0884, 0.2617, 0.2478, 0.1252, 0.0948, 0.1304, 0.0991, 0.1385], + device='cuda:1'), in_proj_covar=tensor([0.0261, 0.0392, 0.0297, 0.0311, 0.0226, 0.0367, 0.0229, 0.0274], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:29:55,767 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.29 vs. limit=5.0 +2023-03-21 09:29:57,111 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6003, 1.8416, 1.5956, 1.6695, 1.8753, 1.7568, 1.8076, 1.3640], + device='cuda:1'), covar=tensor([0.0158, 0.0158, 0.0302, 0.0195, 0.0136, 0.0135, 0.0108, 0.0203], + device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0036, 0.0034, 0.0036, 0.0035, 0.0033, 0.0037, 0.0044], + device='cuda:1'), out_proj_covar=tensor([4.1390e-05, 3.9900e-05, 3.9131e-05, 3.9960e-05, 3.8897e-05, 3.6699e-05, + 4.1211e-05, 4.9121e-05], device='cuda:1') +2023-03-21 09:29:58,027 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=109495.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:30:00,594 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6175, 3.5648, 2.7614, 3.3037, 2.4265, 2.1805, 1.8851, 3.6110], + device='cuda:1'), covar=tensor([0.0057, 0.0062, 0.0205, 0.0097, 0.0252, 0.0593, 0.0661, 0.0063], + device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0095, 0.0114, 0.0095, 0.0130, 0.0137, 0.0130, 0.0105], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 09:30:07,006 INFO [train.py:901] (1/2) Epoch 39, batch 2200, loss[loss=0.1302, simple_loss=0.2083, pruned_loss=0.02607, over 7339.00 frames. ], tot_loss[loss=0.1307, simple_loss=0.2118, pruned_loss=0.02483, over 1441002.36 frames. ], batch size: 61, lr: 4.22e-03, grad_scale: 8.0 +2023-03-21 09:30:07,290 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.85 vs. limit=5.0 +2023-03-21 09:30:11,540 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 09:30:26,280 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.848e+02 2.118e+02 2.463e+02 5.500e+02, threshold=4.236e+02, percent-clipped=2.0 +2023-03-21 09:30:26,389 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109550.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:30:33,357 INFO [train.py:901] (1/2) Epoch 39, batch 2250, loss[loss=0.1433, simple_loss=0.2246, pruned_loss=0.03094, over 7293.00 frames. ], tot_loss[loss=0.1309, simple_loss=0.2122, pruned_loss=0.02478, over 1442058.10 frames. ], batch size: 57, lr: 4.22e-03, grad_scale: 8.0 +2023-03-21 09:30:41,441 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109579.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:30:46,661 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 09:30:46,671 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 09:30:58,619 INFO [train.py:901] (1/2) Epoch 39, batch 2300, loss[loss=0.1292, simple_loss=0.2121, pruned_loss=0.02316, over 7344.00 frames. ], tot_loss[loss=0.1312, simple_loss=0.2124, pruned_loss=0.02494, over 1444185.58 frames. ], batch size: 61, lr: 4.22e-03, grad_scale: 8.0 +2023-03-21 09:30:58,632 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 09:31:13,041 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109640.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:31:18,542 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.188e+02 1.786e+02 2.016e+02 2.497e+02 7.057e+02, threshold=4.032e+02, percent-clipped=1.0 +2023-03-21 09:31:18,682 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109650.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:31:18,708 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7896, 2.8072, 3.6169, 3.7083, 3.7628, 3.7626, 3.7388, 3.6907], + device='cuda:1'), covar=tensor([0.0030, 0.0154, 0.0039, 0.0039, 0.0035, 0.0034, 0.0042, 0.0054], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0070, 0.0058, 0.0057, 0.0055, 0.0061, 0.0049, 0.0078], + device='cuda:1'), out_proj_covar=tensor([8.2100e-05, 1.4185e-04, 1.0548e-04, 9.7941e-05, 9.4510e-05, 1.0770e-04, + 9.2773e-05, 1.4558e-04], device='cuda:1') +2023-03-21 09:31:23,127 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109659.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:31:25,030 INFO [train.py:901] (1/2) Epoch 39, batch 2350, loss[loss=0.149, simple_loss=0.2352, pruned_loss=0.03134, over 6757.00 frames. ], tot_loss[loss=0.1311, simple_loss=0.2124, pruned_loss=0.02487, over 1445036.25 frames. ], batch size: 106, lr: 4.21e-03, grad_scale: 8.0 +2023-03-21 09:31:30,662 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109674.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 09:31:37,001 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-03-21 09:31:40,231 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109693.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 09:31:47,036 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 09:31:50,622 INFO [train.py:901] (1/2) Epoch 39, batch 2400, loss[loss=0.1368, simple_loss=0.2202, pruned_loss=0.02668, over 7297.00 frames. ], tot_loss[loss=0.1308, simple_loss=0.2123, pruned_loss=0.02471, over 1443782.85 frames. ], batch size: 68, lr: 4.21e-03, grad_scale: 8.0 +2023-03-21 09:31:53,143 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 09:31:54,290 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109720.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:32:03,821 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 09:32:05,352 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=109741.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 09:32:06,298 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 09:32:09,798 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.181e+02 1.635e+02 2.013e+02 2.308e+02 3.868e+02, threshold=4.027e+02, percent-clipped=0.0 +2023-03-21 09:32:16,318 INFO [train.py:901] (1/2) Epoch 39, batch 2450, loss[loss=0.1331, simple_loss=0.2158, pruned_loss=0.02521, over 7317.00 frames. ], tot_loss[loss=0.1308, simple_loss=0.2122, pruned_loss=0.02472, over 1442593.31 frames. ], batch size: 59, lr: 4.21e-03, grad_scale: 8.0 +2023-03-21 09:32:23,548 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 +2023-03-21 09:32:32,778 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 09:32:40,374 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9697, 3.3009, 3.9912, 4.0324, 4.0236, 4.0091, 4.0626, 3.9261], + device='cuda:1'), covar=tensor([0.0032, 0.0112, 0.0032, 0.0032, 0.0028, 0.0034, 0.0029, 0.0050], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0068, 0.0057, 0.0055, 0.0054, 0.0060, 0.0048, 0.0077], + device='cuda:1'), out_proj_covar=tensor([8.0188e-05, 1.3750e-04, 1.0281e-04, 9.5632e-05, 9.2117e-05, 1.0517e-04, + 9.0075e-05, 1.4227e-04], device='cuda:1') +2023-03-21 09:32:42,310 INFO [train.py:901] (1/2) Epoch 39, batch 2500, loss[loss=0.09696, simple_loss=0.1741, pruned_loss=0.009906, over 7033.00 frames. ], tot_loss[loss=0.1306, simple_loss=0.2117, pruned_loss=0.02468, over 1440983.70 frames. ], batch size: 35, lr: 4.21e-03, grad_scale: 8.0 +2023-03-21 09:32:48,046 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5571, 2.4855, 2.4617, 3.8350, 1.8792, 3.5886, 1.4440, 3.4879], + device='cuda:1'), covar=tensor([0.0157, 0.1405, 0.1784, 0.0199, 0.4046, 0.0311, 0.1283, 0.0397], + device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0245, 0.0259, 0.0206, 0.0249, 0.0216, 0.0225, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:32:58,342 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 09:33:01,308 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.761e+02 2.069e+02 2.509e+02 3.938e+02, threshold=4.137e+02, percent-clipped=0.0 +2023-03-21 09:33:01,459 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109850.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:33:02,958 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2549, 3.8767, 3.8092, 3.9757, 3.8407, 3.7102, 4.0712, 3.6210], + device='cuda:1'), covar=tensor([0.0120, 0.0177, 0.0155, 0.0172, 0.0463, 0.0140, 0.0155, 0.0185], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0105, 0.0106, 0.0092, 0.0183, 0.0111, 0.0108, 0.0117], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:33:07,855 INFO [train.py:901] (1/2) Epoch 39, batch 2550, loss[loss=0.1293, simple_loss=0.214, pruned_loss=0.02231, over 7326.00 frames. ], tot_loss[loss=0.1306, simple_loss=0.2116, pruned_loss=0.02479, over 1440652.98 frames. ], batch size: 75, lr: 4.21e-03, grad_scale: 16.0 +2023-03-21 09:33:07,983 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7252, 2.8227, 3.6762, 3.6756, 3.7494, 3.7030, 3.5909, 3.6308], + device='cuda:1'), covar=tensor([0.0029, 0.0142, 0.0035, 0.0036, 0.0031, 0.0035, 0.0045, 0.0053], + device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0068, 0.0056, 0.0055, 0.0054, 0.0060, 0.0048, 0.0077], + device='cuda:1'), out_proj_covar=tensor([8.0249e-05, 1.3745e-04, 1.0253e-04, 9.5538e-05, 9.2024e-05, 1.0524e-04, + 8.9790e-05, 1.4205e-04], device='cuda:1') +2023-03-21 09:33:19,677 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2523, 2.1518, 2.3826, 2.1265, 2.4088, 2.2678, 2.0261, 1.7290], + device='cuda:1'), covar=tensor([0.0471, 0.0608, 0.0350, 0.0298, 0.0344, 0.0441, 0.0398, 0.0337], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0039, 0.0040, 0.0039, 0.0037, 0.0037, 0.0042, 0.0042], + device='cuda:1'), out_proj_covar=tensor([1.0139e-04, 1.0172e-04, 1.0138e-04, 9.8930e-05, 9.6773e-05, 9.7138e-05, + 1.0522e-04, 1.0626e-04], device='cuda:1') +2023-03-21 09:33:26,174 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=109898.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:33:34,171 INFO [train.py:901] (1/2) Epoch 39, batch 2600, loss[loss=0.1412, simple_loss=0.2264, pruned_loss=0.02799, over 7231.00 frames. ], tot_loss[loss=0.1302, simple_loss=0.2112, pruned_loss=0.02456, over 1439249.63 frames. ], batch size: 93, lr: 4.21e-03, grad_scale: 8.0 +2023-03-21 09:33:45,219 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109935.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:33:52,349 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 +2023-03-21 09:33:52,612 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109950.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:33:52,979 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.178e+02 1.771e+02 2.022e+02 2.327e+02 3.789e+02, threshold=4.044e+02, percent-clipped=0.0 +2023-03-21 09:33:58,875 INFO [train.py:901] (1/2) Epoch 39, batch 2650, loss[loss=0.1379, simple_loss=0.2177, pruned_loss=0.02904, over 7280.00 frames. ], tot_loss[loss=0.1303, simple_loss=0.2112, pruned_loss=0.02467, over 1439589.17 frames. ], batch size: 57, lr: 4.21e-03, grad_scale: 8.0 +2023-03-21 09:34:04,498 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109974.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:34:16,343 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=109998.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:34:23,965 INFO [train.py:901] (1/2) Epoch 39, batch 2700, loss[loss=0.1067, simple_loss=0.1783, pruned_loss=0.01752, over 6948.00 frames. ], tot_loss[loss=0.1299, simple_loss=0.211, pruned_loss=0.0244, over 1441528.18 frames. ], batch size: 35, lr: 4.21e-03, grad_scale: 8.0 +2023-03-21 09:34:25,034 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110015.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:34:26,105 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110017.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:34:28,518 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110022.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:34:31,548 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7946, 2.4093, 3.0239, 2.8491, 2.8822, 2.8363, 2.3789, 2.8720], + device='cuda:1'), covar=tensor([0.1407, 0.0776, 0.0761, 0.0924, 0.0871, 0.1015, 0.1966, 0.1216], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0067, 0.0050, 0.0050, 0.0049, 0.0048, 0.0067, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 09:34:39,334 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110044.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:34:42,666 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.192e+02 1.798e+02 2.201e+02 2.521e+02 4.025e+02, threshold=4.402e+02, percent-clipped=0.0 +2023-03-21 09:34:48,452 INFO [train.py:901] (1/2) Epoch 39, batch 2750, loss[loss=0.1437, simple_loss=0.2188, pruned_loss=0.03428, over 7197.00 frames. ], tot_loss[loss=0.1303, simple_loss=0.2113, pruned_loss=0.02467, over 1441418.20 frames. ], batch size: 45, lr: 4.21e-03, grad_scale: 8.0 +2023-03-21 09:34:55,955 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110078.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:35:09,647 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110105.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:35:12,053 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110110.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:35:13,408 INFO [train.py:901] (1/2) Epoch 39, batch 2800, loss[loss=0.1114, simple_loss=0.1979, pruned_loss=0.01248, over 7107.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.2107, pruned_loss=0.0243, over 1442309.06 frames. ], batch size: 41, lr: 4.21e-03, grad_scale: 8.0 +2023-03-21 09:35:14,465 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110115.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:35:38,122 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 09:35:38,172 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 09:35:38,196 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 09:35:38,504 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 09:35:38,672 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 09:35:38,689 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 09:35:38,701 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 09:35:38,798 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 09:35:38,870 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 09:35:38,885 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 09:35:38,906 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 09:35:39,066 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 09:35:39,072 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 09:35:39,079 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 09:35:39,509 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 09:35:39,511 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 09:35:39,568 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 09:35:45,300 INFO [train.py:901] (1/2) Epoch 40, batch 0, loss[loss=0.1322, simple_loss=0.2162, pruned_loss=0.02412, over 7322.00 frames. ], tot_loss[loss=0.1322, simple_loss=0.2162, pruned_loss=0.02412, over 7322.00 frames. ], batch size: 75, lr: 4.15e-03, grad_scale: 8.0 +2023-03-21 09:35:45,300 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 09:35:52,684 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1313, 2.7143, 3.2168, 3.1466, 2.9591, 2.8729, 3.0717, 2.3752], + device='cuda:1'), covar=tensor([0.0373, 0.0389, 0.0723, 0.0584, 0.0632, 0.0929, 0.0594, 0.2397], + device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0336, 0.0273, 0.0357, 0.0289, 0.0288, 0.0348, 0.0248], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:35:53,427 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0429, 2.9814, 2.2898, 3.4399, 2.3455, 2.8121, 1.5702, 2.4614], + device='cuda:1'), covar=tensor([0.0526, 0.0758, 0.2684, 0.0734, 0.0419, 0.0472, 0.4055, 0.1670], + device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0258, 0.0283, 0.0270, 0.0271, 0.0264, 0.0234, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:36:11,282 INFO [train.py:935] (1/2) Epoch 40, validation: loss=0.1653, simple_loss=0.2572, pruned_loss=0.03673, over 1622729.00 frames. +2023-03-21 09:36:11,283 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 09:36:18,228 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 09:36:18,683 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.279e+02 1.747e+02 2.118e+02 2.577e+02 4.675e+02, threshold=4.235e+02, percent-clipped=1.0 +2023-03-21 09:36:28,752 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 09:36:28,896 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110171.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:36:31,354 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110176.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:36:36,124 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 09:36:36,614 INFO [train.py:901] (1/2) Epoch 40, batch 50, loss[loss=0.1314, simple_loss=0.2139, pruned_loss=0.02445, over 7360.00 frames. ], tot_loss[loss=0.1328, simple_loss=0.2138, pruned_loss=0.02593, over 325469.32 frames. ], batch size: 65, lr: 4.15e-03, grad_scale: 8.0 +2023-03-21 09:36:38,626 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 09:36:41,608 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 09:36:47,131 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1497, 3.2298, 3.6018, 3.3387, 3.4099, 3.3262, 2.8079, 3.2754], + device='cuda:1'), covar=tensor([0.1566, 0.0641, 0.0804, 0.1263, 0.0782, 0.0952, 0.1783, 0.1388], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0068, 0.0051, 0.0051, 0.0050, 0.0048, 0.0068, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 09:36:59,059 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 09:36:59,532 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 09:37:02,118 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110235.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:37:02,970 INFO [train.py:901] (1/2) Epoch 40, batch 100, loss[loss=0.1296, simple_loss=0.214, pruned_loss=0.02258, over 7285.00 frames. ], tot_loss[loss=0.1291, simple_loss=0.2104, pruned_loss=0.02389, over 573079.90 frames. ], batch size: 66, lr: 4.15e-03, grad_scale: 8.0 +2023-03-21 09:37:09,490 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-03-21 09:37:10,703 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.749e+02 2.051e+02 2.351e+02 3.808e+02, threshold=4.103e+02, percent-clipped=0.0 +2023-03-21 09:37:26,981 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:37:27,069 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:37:28,119 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0182, 2.7569, 3.0734, 3.0489, 2.7309, 2.7057, 3.0569, 2.2541], + device='cuda:1'), covar=tensor([0.0417, 0.0490, 0.0670, 0.0655, 0.0686, 0.0956, 0.0641, 0.2382], + device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0335, 0.0271, 0.0355, 0.0288, 0.0287, 0.0345, 0.0247], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:37:28,929 INFO [train.py:901] (1/2) Epoch 40, batch 150, loss[loss=0.1373, simple_loss=0.2159, pruned_loss=0.0294, over 7273.00 frames. ], tot_loss[loss=0.1295, simple_loss=0.2109, pruned_loss=0.02404, over 765917.04 frames. ], batch size: 52, lr: 4.15e-03, grad_scale: 8.0 +2023-03-21 09:37:37,090 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110303.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:37:43,756 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110315.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:37:45,362 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110318.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:37:55,291 INFO [train.py:901] (1/2) Epoch 40, batch 200, loss[loss=0.1248, simple_loss=0.2008, pruned_loss=0.02447, over 7287.00 frames. ], tot_loss[loss=0.13, simple_loss=0.2112, pruned_loss=0.02441, over 913491.37 frames. ], batch size: 77, lr: 4.15e-03, grad_scale: 8.0 +2023-03-21 09:37:59,024 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110344.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:38:00,961 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 09:38:02,481 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.033e+02 1.699e+02 2.011e+02 2.332e+02 3.842e+02, threshold=4.021e+02, percent-clipped=0.0 +2023-03-21 09:38:05,525 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 09:38:08,569 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110363.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:38:09,132 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110364.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:38:10,977 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 09:38:13,527 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110373.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:38:16,679 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110379.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:38:20,509 INFO [train.py:901] (1/2) Epoch 40, batch 250, loss[loss=0.1442, simple_loss=0.2193, pruned_loss=0.03451, over 7318.00 frames. ], tot_loss[loss=0.1304, simple_loss=0.2117, pruned_loss=0.02452, over 1030660.77 frames. ], batch size: 59, lr: 4.15e-03, grad_scale: 8.0 +2023-03-21 09:38:22,991 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 09:38:28,095 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110400.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:38:45,424 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 09:38:46,964 INFO [train.py:901] (1/2) Epoch 40, batch 300, loss[loss=0.09994, simple_loss=0.1825, pruned_loss=0.008712, over 7292.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.2109, pruned_loss=0.0242, over 1122912.28 frames. ], batch size: 42, lr: 4.15e-03, grad_scale: 8.0 +2023-03-21 09:38:53,520 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1182, 3.1675, 2.3445, 3.5246, 2.6910, 2.8480, 1.6449, 2.5292], + device='cuda:1'), covar=tensor([0.0440, 0.0801, 0.2452, 0.0686, 0.0424, 0.0572, 0.3690, 0.1602], + device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0258, 0.0281, 0.0270, 0.0270, 0.0265, 0.0234, 0.0257], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:38:53,832 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 1.777e+02 2.048e+02 2.389e+02 4.929e+02, threshold=4.095e+02, percent-clipped=1.0 +2023-03-21 09:38:53,878 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 09:38:59,977 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5190, 2.3048, 2.4258, 3.6924, 1.8474, 3.4501, 1.3191, 3.3689], + device='cuda:1'), covar=tensor([0.0189, 0.1505, 0.1902, 0.0246, 0.4001, 0.0251, 0.1373, 0.0455], + device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0245, 0.0262, 0.0207, 0.0251, 0.0216, 0.0226, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:39:01,363 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110466.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:39:03,918 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110471.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:39:08,945 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4271, 1.7157, 1.6244, 1.5213, 1.6611, 1.5662, 1.4318, 1.3583], + device='cuda:1'), covar=tensor([0.0124, 0.0155, 0.0168, 0.0147, 0.0114, 0.0142, 0.0157, 0.0189], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0037, 0.0035, 0.0038, 0.0036, 0.0034, 0.0038, 0.0046], + device='cuda:1'), out_proj_covar=tensor([4.2461e-05, 4.0772e-05, 3.9854e-05, 4.1526e-05, 4.0017e-05, 3.7816e-05, + 4.2304e-05, 5.0383e-05], device='cuda:1') +2023-03-21 09:39:12,426 INFO [train.py:901] (1/2) Epoch 40, batch 350, loss[loss=0.1031, simple_loss=0.1747, pruned_loss=0.01578, over 6975.00 frames. ], tot_loss[loss=0.1294, simple_loss=0.2107, pruned_loss=0.02407, over 1195608.53 frames. ], batch size: 35, lr: 4.15e-03, grad_scale: 8.0 +2023-03-21 09:39:17,774 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-03-21 09:39:21,346 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 +2023-03-21 09:39:28,536 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 09:39:37,974 INFO [train.py:901] (1/2) Epoch 40, batch 400, loss[loss=0.134, simple_loss=0.2185, pruned_loss=0.02474, over 7285.00 frames. ], tot_loss[loss=0.1298, simple_loss=0.211, pruned_loss=0.02432, over 1249865.93 frames. ], batch size: 66, lr: 4.14e-03, grad_scale: 8.0 +2023-03-21 09:39:45,023 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.773e+02 2.034e+02 2.439e+02 3.776e+02, threshold=4.069e+02, percent-clipped=0.0 +2023-03-21 09:39:45,164 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0897, 2.8922, 3.4064, 2.9410, 3.2380, 3.0123, 2.7152, 2.9289], + device='cuda:1'), covar=tensor([0.1555, 0.0609, 0.0750, 0.1744, 0.0727, 0.1030, 0.1807, 0.2004], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0067, 0.0051, 0.0051, 0.0050, 0.0049, 0.0068, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 09:39:56,798 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6339, 2.0409, 1.8148, 1.8183, 2.0443, 1.8115, 1.8086, 1.5429], + device='cuda:1'), covar=tensor([0.0139, 0.0242, 0.0273, 0.0276, 0.0122, 0.0137, 0.0180, 0.0279], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0036, 0.0035, 0.0037, 0.0036, 0.0034, 0.0038, 0.0045], + device='cuda:1'), out_proj_covar=tensor([4.2246e-05, 4.0460e-05, 3.9637e-05, 4.1261e-05, 3.9759e-05, 3.7537e-05, + 4.2247e-05, 5.0178e-05], device='cuda:1') +2023-03-21 09:40:04,285 INFO [train.py:901] (1/2) Epoch 40, batch 450, loss[loss=0.1524, simple_loss=0.2419, pruned_loss=0.03141, over 6680.00 frames. ], tot_loss[loss=0.1303, simple_loss=0.2118, pruned_loss=0.02438, over 1292840.52 frames. ], batch size: 106, lr: 4.14e-03, grad_scale: 8.0 +2023-03-21 09:40:08,883 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 09:40:09,889 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 09:40:16,900 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110611.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:40:24,199 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 +2023-03-21 09:40:29,884 INFO [train.py:901] (1/2) Epoch 40, batch 500, loss[loss=0.1084, simple_loss=0.1881, pruned_loss=0.01441, over 6922.00 frames. ], tot_loss[loss=0.1301, simple_loss=0.2117, pruned_loss=0.02427, over 1326729.63 frames. ], batch size: 35, lr: 4.14e-03, grad_scale: 8.0 +2023-03-21 09:40:30,971 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110639.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:40:36,902 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.326e+02 1.687e+02 1.978e+02 2.331e+02 4.016e+02, threshold=3.957e+02, percent-clipped=0.0 +2023-03-21 09:40:41,568 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110659.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:40:43,066 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 09:40:44,616 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 09:40:45,119 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 09:40:47,123 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 09:40:48,808 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110672.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:40:49,241 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110673.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:40:49,733 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110674.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:40:51,679 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 09:40:56,160 INFO [train.py:901] (1/2) Epoch 40, batch 550, loss[loss=0.09568, simple_loss=0.159, pruned_loss=0.01618, over 6040.00 frames. ], tot_loss[loss=0.1299, simple_loss=0.2112, pruned_loss=0.02431, over 1352467.67 frames. ], batch size: 26, lr: 4.14e-03, grad_scale: 8.0 +2023-03-21 09:41:02,697 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 09:41:02,774 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110700.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:41:05,996 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.46 vs. limit=5.0 +2023-03-21 09:41:06,318 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9321, 2.5572, 3.0515, 2.9469, 3.0614, 2.7517, 2.6256, 2.8203], + device='cuda:1'), covar=tensor([0.1261, 0.0799, 0.1025, 0.1121, 0.0717, 0.1210, 0.1517, 0.1984], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0068, 0.0051, 0.0051, 0.0050, 0.0050, 0.0068, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 09:41:10,797 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4310, 2.5613, 2.6629, 2.1984, 2.5719, 2.4491, 2.1241, 1.8863], + device='cuda:1'), covar=tensor([0.0323, 0.0325, 0.0229, 0.0278, 0.0498, 0.0338, 0.0304, 0.0327], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0038, 0.0039, 0.0038, 0.0036, 0.0036, 0.0041, 0.0041], + device='cuda:1'), out_proj_covar=tensor([9.8369e-05, 9.8673e-05, 9.8501e-05, 9.5905e-05, 9.4125e-05, 9.4368e-05, + 1.0220e-04, 1.0326e-04], device='cuda:1') +2023-03-21 09:41:11,157 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 09:41:13,192 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110721.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:41:14,140 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 09:41:20,858 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110736.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:41:21,225 INFO [train.py:901] (1/2) Epoch 40, batch 600, loss[loss=0.1236, simple_loss=0.206, pruned_loss=0.02056, over 7281.00 frames. ], tot_loss[loss=0.1305, simple_loss=0.2116, pruned_loss=0.02467, over 1372197.89 frames. ], batch size: 77, lr: 4.14e-03, grad_scale: 8.0 +2023-03-21 09:41:21,232 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 09:41:27,438 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110748.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:41:29,433 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.347e+02 1.717e+02 1.997e+02 2.356e+02 3.346e+02, threshold=3.993e+02, percent-clipped=0.0 +2023-03-21 09:41:37,201 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110766.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:41:38,150 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 09:41:39,719 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110771.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:41:46,652 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 09:41:47,608 INFO [train.py:901] (1/2) Epoch 40, batch 650, loss[loss=0.1473, simple_loss=0.2256, pruned_loss=0.03452, over 7328.00 frames. ], tot_loss[loss=0.1302, simple_loss=0.2114, pruned_loss=0.0245, over 1386516.78 frames. ], batch size: 59, lr: 4.14e-03, grad_scale: 8.0 +2023-03-21 09:41:52,670 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110797.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 09:42:01,427 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110814.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:42:03,981 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110819.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:42:04,408 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 09:42:06,092 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4408, 1.7407, 1.4685, 1.5728, 1.7476, 1.6057, 1.5633, 1.4209], + device='cuda:1'), covar=tensor([0.0193, 0.0263, 0.0310, 0.0223, 0.0168, 0.0214, 0.0145, 0.0224], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0037, 0.0035, 0.0038, 0.0036, 0.0034, 0.0037, 0.0046], + device='cuda:1'), out_proj_covar=tensor([4.2782e-05, 4.0703e-05, 3.9932e-05, 4.1408e-05, 3.9877e-05, 3.7571e-05, + 4.2096e-05, 5.0563e-05], device='cuda:1') +2023-03-21 09:42:13,028 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 09:42:14,122 INFO [train.py:901] (1/2) Epoch 40, batch 700, loss[loss=0.1171, simple_loss=0.1939, pruned_loss=0.02018, over 7152.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.2108, pruned_loss=0.02422, over 1400173.60 frames. ], batch size: 41, lr: 4.14e-03, grad_scale: 8.0 +2023-03-21 09:42:21,054 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.283e+02 1.664e+02 1.895e+02 2.126e+02 3.944e+02, threshold=3.791e+02, percent-clipped=0.0 +2023-03-21 09:42:29,780 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110868.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:42:38,143 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 09:42:38,640 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 09:42:39,152 INFO [train.py:901] (1/2) Epoch 40, batch 750, loss[loss=0.1358, simple_loss=0.219, pruned_loss=0.02627, over 7287.00 frames. ], tot_loss[loss=0.1297, simple_loss=0.2107, pruned_loss=0.02428, over 1409771.82 frames. ], batch size: 66, lr: 4.14e-03, grad_scale: 8.0 +2023-03-21 09:42:40,801 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3742, 1.6576, 1.5045, 1.5882, 1.7652, 1.6357, 1.6105, 1.3160], + device='cuda:1'), covar=tensor([0.0174, 0.0213, 0.0293, 0.0157, 0.0158, 0.0110, 0.0179, 0.0213], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0037, 0.0035, 0.0038, 0.0036, 0.0034, 0.0038, 0.0046], + device='cuda:1'), out_proj_covar=tensor([4.3161e-05, 4.0774e-05, 4.0041e-05, 4.1683e-05, 4.0167e-05, 3.7726e-05, + 4.2311e-05, 5.0895e-05], device='cuda:1') +2023-03-21 09:42:51,528 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 09:42:56,686 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 09:43:01,394 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110929.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:43:02,785 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 09:43:04,295 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 09:43:05,220 INFO [train.py:901] (1/2) Epoch 40, batch 800, loss[loss=0.1496, simple_loss=0.2392, pruned_loss=0.03001, over 7317.00 frames. ], tot_loss[loss=0.1302, simple_loss=0.2114, pruned_loss=0.0245, over 1415524.68 frames. ], batch size: 61, lr: 4.14e-03, grad_scale: 8.0 +2023-03-21 09:43:06,289 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110939.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:43:12,147 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.823e+02 2.168e+02 2.507e+02 3.727e+02, threshold=4.336e+02, percent-clipped=0.0 +2023-03-21 09:43:14,259 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 09:43:15,932 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110958.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:43:16,367 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110959.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:43:20,324 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110967.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:43:23,951 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110974.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:43:25,369 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1952, 4.6637, 4.7532, 4.7327, 4.6575, 4.2094, 4.7778, 4.6048], + device='cuda:1'), covar=tensor([0.0517, 0.0439, 0.0405, 0.0441, 0.0406, 0.0499, 0.0388, 0.0573], + device='cuda:1'), in_proj_covar=tensor([0.0220, 0.0262, 0.0204, 0.0203, 0.0163, 0.0232, 0.0212, 0.0151], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:43:30,313 INFO [train.py:901] (1/2) Epoch 40, batch 850, loss[loss=0.1238, simple_loss=0.203, pruned_loss=0.02233, over 7367.00 frames. ], tot_loss[loss=0.1298, simple_loss=0.2111, pruned_loss=0.0243, over 1422279.94 frames. ], batch size: 44, lr: 4.14e-03, grad_scale: 8.0 +2023-03-21 09:43:30,361 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=110987.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:43:30,965 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3379, 1.6356, 1.3971, 1.4517, 1.5650, 1.5501, 1.5057, 1.3233], + device='cuda:1'), covar=tensor([0.0166, 0.0169, 0.0197, 0.0160, 0.0132, 0.0130, 0.0130, 0.0179], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0036, 0.0035, 0.0037, 0.0036, 0.0034, 0.0038, 0.0046], + device='cuda:1'), out_proj_covar=tensor([4.3072e-05, 4.0443e-05, 3.9814e-05, 4.1336e-05, 3.9875e-05, 3.7547e-05, + 4.2224e-05, 5.0570e-05], device='cuda:1') +2023-03-21 09:43:33,293 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 09:43:33,301 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 09:43:39,735 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 09:43:41,248 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=111007.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:43:41,906 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3967, 3.4509, 2.4878, 3.8532, 2.7383, 3.3819, 1.7473, 2.5911], + device='cuda:1'), covar=tensor([0.0540, 0.0748, 0.2798, 0.0696, 0.0526, 0.0736, 0.3903, 0.1819], + device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0259, 0.0284, 0.0272, 0.0272, 0.0267, 0.0235, 0.0259], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:43:43,249 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 09:43:47,946 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2356, 2.3129, 2.4550, 2.1766, 2.4242, 2.3249, 1.9587, 1.7256], + device='cuda:1'), covar=tensor([0.0413, 0.0370, 0.0398, 0.0232, 0.0421, 0.0389, 0.0436, 0.0458], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0039, 0.0040, 0.0039, 0.0037, 0.0037, 0.0042, 0.0042], + device='cuda:1'), out_proj_covar=tensor([1.0175e-04, 1.0177e-04, 1.0110e-04, 9.8813e-05, 9.7296e-05, 9.7278e-05, + 1.0509e-04, 1.0630e-04], device='cuda:1') +2023-03-21 09:43:47,972 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111019.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:43:49,374 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=111022.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:43:51,465 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3827, 2.5068, 2.7755, 2.3225, 2.7674, 2.3835, 2.1888, 2.0077], + device='cuda:1'), covar=tensor([0.0778, 0.0721, 0.0497, 0.0301, 0.0593, 0.0507, 0.0425, 0.0443], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0039, 0.0040, 0.0039, 0.0037, 0.0038, 0.0042, 0.0042], + device='cuda:1'), out_proj_covar=tensor([1.0186e-04, 1.0191e-04, 1.0127e-04, 9.8997e-05, 9.7532e-05, 9.7462e-05, + 1.0533e-04, 1.0647e-04], device='cuda:1') +2023-03-21 09:43:56,763 INFO [train.py:901] (1/2) Epoch 40, batch 900, loss[loss=0.12, simple_loss=0.2014, pruned_loss=0.01925, over 7164.00 frames. ], tot_loss[loss=0.1302, simple_loss=0.2113, pruned_loss=0.02449, over 1424997.11 frames. ], batch size: 41, lr: 4.14e-03, grad_scale: 8.0 +2023-03-21 09:44:03,877 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.196e+02 1.783e+02 2.026e+02 2.389e+02 5.093e+02, threshold=4.052e+02, percent-clipped=1.0 +2023-03-21 09:44:07,657 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1680, 3.1627, 2.3089, 3.4935, 2.3383, 3.0585, 1.5474, 2.4155], + device='cuda:1'), covar=tensor([0.0639, 0.0966, 0.2758, 0.0838, 0.0565, 0.0670, 0.4032, 0.1841], + device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0260, 0.0284, 0.0272, 0.0273, 0.0268, 0.0236, 0.0260], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:44:20,191 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 09:44:22,198 INFO [train.py:901] (1/2) Epoch 40, batch 950, loss[loss=0.1275, simple_loss=0.2048, pruned_loss=0.02505, over 7210.00 frames. ], tot_loss[loss=0.1299, simple_loss=0.211, pruned_loss=0.02442, over 1427345.68 frames. ], batch size: 50, lr: 4.13e-03, grad_scale: 8.0 +2023-03-21 09:44:25,428 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111092.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 09:44:30,537 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3894, 4.8836, 4.9199, 4.8995, 4.8112, 4.4468, 4.9450, 4.7982], + device='cuda:1'), covar=tensor([0.0463, 0.0376, 0.0383, 0.0399, 0.0317, 0.0368, 0.0330, 0.0416], + device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0259, 0.0202, 0.0201, 0.0162, 0.0230, 0.0211, 0.0149], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:44:45,181 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 09:44:48,696 INFO [train.py:901] (1/2) Epoch 40, batch 1000, loss[loss=0.1629, simple_loss=0.2461, pruned_loss=0.03988, over 6662.00 frames. ], tot_loss[loss=0.1295, simple_loss=0.2106, pruned_loss=0.02419, over 1431282.80 frames. ], batch size: 106, lr: 4.13e-03, grad_scale: 8.0 +2023-03-21 09:44:55,152 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 +2023-03-21 09:44:55,751 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.230e+02 1.691e+02 1.890e+02 2.255e+02 3.563e+02, threshold=3.780e+02, percent-clipped=0.0 +2023-03-21 09:45:05,715 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 09:45:14,424 INFO [train.py:901] (1/2) Epoch 40, batch 1050, loss[loss=0.1422, simple_loss=0.2191, pruned_loss=0.03266, over 7198.00 frames. ], tot_loss[loss=0.1298, simple_loss=0.2109, pruned_loss=0.02438, over 1434415.98 frames. ], batch size: 50, lr: 4.13e-03, grad_scale: 8.0 +2023-03-21 09:45:21,193 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8527, 3.0863, 2.8078, 2.9721, 3.1669, 2.6430, 3.0736, 2.8950], + device='cuda:1'), covar=tensor([0.0671, 0.0785, 0.0894, 0.1130, 0.0656, 0.0729, 0.0722, 0.1317], + device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0059, 0.0067, 0.0060, 0.0057, 0.0062, 0.0057, 0.0054], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:45:29,295 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 09:45:32,529 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6615, 1.5594, 1.8564, 2.1329, 1.9328, 2.0834, 1.5483, 2.0384], + device='cuda:1'), covar=tensor([0.2075, 0.4647, 0.1288, 0.1151, 0.2269, 0.1236, 0.2130, 0.1761], + device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0078, 0.0071, 0.0063, 0.0064, 0.0063, 0.0104, 0.0068], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:45:33,421 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 09:45:33,998 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111224.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:45:34,563 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111225.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:45:40,518 INFO [train.py:901] (1/2) Epoch 40, batch 1100, loss[loss=0.1214, simple_loss=0.1967, pruned_loss=0.02303, over 7158.00 frames. ], tot_loss[loss=0.1299, simple_loss=0.211, pruned_loss=0.02436, over 1434093.60 frames. ], batch size: 39, lr: 4.13e-03, grad_scale: 8.0 +2023-03-21 09:45:47,522 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 1.688e+02 1.957e+02 2.307e+02 4.408e+02, threshold=3.914e+02, percent-clipped=3.0 +2023-03-21 09:45:55,911 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6405, 3.7175, 3.5090, 3.6376, 3.4011, 3.5662, 3.9612, 3.9968], + device='cuda:1'), covar=tensor([0.0211, 0.0170, 0.0287, 0.0208, 0.0344, 0.0504, 0.0234, 0.0189], + device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0130, 0.0123, 0.0126, 0.0116, 0.0103, 0.0100, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:45:56,438 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111267.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:45:59,836 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-03-21 09:46:02,000 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 09:46:02,011 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 09:46:03,162 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9790, 2.4421, 1.8521, 2.6709, 2.6055, 2.4389, 2.6259, 2.5274], + device='cuda:1'), covar=tensor([0.2281, 0.1270, 0.4088, 0.0668, 0.0334, 0.0295, 0.0466, 0.0369], + device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0227, 0.0244, 0.0256, 0.0196, 0.0196, 0.0216, 0.0222], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:46:06,600 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111286.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:46:06,950 INFO [train.py:901] (1/2) Epoch 40, batch 1150, loss[loss=0.1072, simple_loss=0.1754, pruned_loss=0.01945, over 7022.00 frames. ], tot_loss[loss=0.1303, simple_loss=0.2115, pruned_loss=0.02454, over 1434365.09 frames. ], batch size: 35, lr: 4.13e-03, grad_scale: 8.0 +2023-03-21 09:46:14,993 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 09:46:15,498 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 09:46:20,493 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111314.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:46:20,988 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=111315.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:46:32,021 INFO [train.py:901] (1/2) Epoch 40, batch 1200, loss[loss=0.1297, simple_loss=0.2123, pruned_loss=0.02352, over 7340.00 frames. ], tot_loss[loss=0.1304, simple_loss=0.2116, pruned_loss=0.02459, over 1435344.58 frames. ], batch size: 61, lr: 4.13e-03, grad_scale: 8.0 +2023-03-21 09:46:36,080 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4327, 4.1919, 3.9539, 4.0577, 3.5277, 2.4938, 2.0617, 4.3653], + device='cuda:1'), covar=tensor([0.0047, 0.0078, 0.0094, 0.0059, 0.0148, 0.0602, 0.0670, 0.0061], + device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0096, 0.0113, 0.0094, 0.0132, 0.0137, 0.0130, 0.0106], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 09:46:39,639 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.301e+02 1.737e+02 2.081e+02 2.449e+02 8.398e+02, threshold=4.161e+02, percent-clipped=1.0 +2023-03-21 09:46:47,694 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 09:46:58,231 INFO [train.py:901] (1/2) Epoch 40, batch 1250, loss[loss=0.157, simple_loss=0.2333, pruned_loss=0.0404, over 7324.00 frames. ], tot_loss[loss=0.1305, simple_loss=0.2119, pruned_loss=0.02453, over 1437780.89 frames. ], batch size: 54, lr: 4.13e-03, grad_scale: 8.0 +2023-03-21 09:47:00,832 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111392.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 09:47:11,203 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 09:47:15,166 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 09:47:16,199 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 09:47:24,390 INFO [train.py:901] (1/2) Epoch 40, batch 1300, loss[loss=0.1262, simple_loss=0.2034, pruned_loss=0.02452, over 7246.00 frames. ], tot_loss[loss=0.13, simple_loss=0.2113, pruned_loss=0.02429, over 1437175.68 frames. ], batch size: 47, lr: 4.13e-03, grad_scale: 8.0 +2023-03-21 09:47:26,553 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=111440.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:47:30,577 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9189, 2.4484, 3.0961, 2.8861, 2.9617, 2.7594, 2.5380, 3.0955], + device='cuda:1'), covar=tensor([0.1333, 0.0852, 0.0911, 0.0988, 0.0716, 0.1025, 0.1568, 0.0844], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0066, 0.0050, 0.0049, 0.0049, 0.0048, 0.0067, 0.0050], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 09:47:31,940 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.074e+02 1.753e+02 2.042e+02 2.404e+02 3.175e+02, threshold=4.084e+02, percent-clipped=0.0 +2023-03-21 09:47:39,667 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 09:47:40,839 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111468.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:47:41,742 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 09:47:45,345 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 09:47:50,250 INFO [train.py:901] (1/2) Epoch 40, batch 1350, loss[loss=0.1395, simple_loss=0.2247, pruned_loss=0.02712, over 7151.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.2111, pruned_loss=0.02406, over 1438643.61 frames. ], batch size: 98, lr: 4.13e-03, grad_scale: 8.0 +2023-03-21 09:47:51,913 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8590, 1.8440, 2.1711, 2.6044, 2.3860, 2.3074, 2.2950, 2.4578], + device='cuda:1'), covar=tensor([0.3480, 0.4116, 0.3238, 0.1663, 0.3229, 0.4995, 0.1849, 0.2640], + device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0078, 0.0070, 0.0063, 0.0064, 0.0063, 0.0103, 0.0068], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:47:54,283 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 09:48:09,998 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111524.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:48:12,531 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111529.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:48:15,891 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1897, 4.3471, 4.0975, 4.2665, 3.9112, 4.3536, 4.6590, 4.7060], + device='cuda:1'), covar=tensor([0.0183, 0.0143, 0.0217, 0.0163, 0.0339, 0.0268, 0.0207, 0.0154], + device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0129, 0.0123, 0.0126, 0.0117, 0.0104, 0.0100, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:48:16,294 INFO [train.py:901] (1/2) Epoch 40, batch 1400, loss[loss=0.1209, simple_loss=0.1996, pruned_loss=0.02108, over 7339.00 frames. ], tot_loss[loss=0.1301, simple_loss=0.2115, pruned_loss=0.02432, over 1438245.05 frames. ], batch size: 44, lr: 4.13e-03, grad_scale: 8.0 +2023-03-21 09:48:17,570 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 +2023-03-21 09:48:23,416 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.668e+02 1.991e+02 2.540e+02 5.106e+02, threshold=3.982e+02, percent-clipped=1.0 +2023-03-21 09:48:27,051 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 09:48:34,174 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=111572.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:48:38,683 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111581.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:48:41,624 INFO [train.py:901] (1/2) Epoch 40, batch 1450, loss[loss=0.1345, simple_loss=0.2097, pruned_loss=0.02965, over 7369.00 frames. ], tot_loss[loss=0.1299, simple_loss=0.2111, pruned_loss=0.02434, over 1439997.93 frames. ], batch size: 51, lr: 4.13e-03, grad_scale: 8.0 +2023-03-21 09:48:46,437 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 +2023-03-21 09:48:50,661 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4073, 2.7929, 2.1729, 3.1274, 3.0243, 3.0798, 2.8943, 2.7284], + device='cuda:1'), covar=tensor([0.1981, 0.0999, 0.3682, 0.0636, 0.0292, 0.0294, 0.0383, 0.0415], + device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0228, 0.0244, 0.0256, 0.0197, 0.0197, 0.0216, 0.0222], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 09:48:51,030 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 09:48:56,887 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111614.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:49:07,258 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 09:49:08,251 INFO [train.py:901] (1/2) Epoch 40, batch 1500, loss[loss=0.1418, simple_loss=0.2234, pruned_loss=0.03017, over 7273.00 frames. ], tot_loss[loss=0.1311, simple_loss=0.2125, pruned_loss=0.02482, over 1440526.46 frames. ], batch size: 52, lr: 4.12e-03, grad_scale: 8.0 +2023-03-21 09:49:15,116 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.685e+02 1.982e+02 2.275e+02 4.883e+02, threshold=3.965e+02, percent-clipped=1.0 +2023-03-21 09:49:20,685 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=111662.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:49:30,635 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 09:49:33,129 INFO [train.py:901] (1/2) Epoch 40, batch 1550, loss[loss=0.1341, simple_loss=0.2164, pruned_loss=0.02587, over 7298.00 frames. ], tot_loss[loss=0.1317, simple_loss=0.2131, pruned_loss=0.02517, over 1441087.77 frames. ], batch size: 86, lr: 4.12e-03, grad_scale: 8.0 +2023-03-21 09:49:40,627 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1446, 2.8621, 3.2107, 3.0693, 2.8688, 2.8224, 3.1997, 2.3095], + device='cuda:1'), covar=tensor([0.0560, 0.0583, 0.0631, 0.0796, 0.0697, 0.1134, 0.0777, 0.2495], + device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0333, 0.0271, 0.0354, 0.0284, 0.0284, 0.0343, 0.0243], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:49:42,829 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-21 09:49:53,251 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 +2023-03-21 09:49:56,075 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6841, 1.9835, 1.7396, 1.8732, 1.9485, 1.7105, 1.9006, 1.4181], + device='cuda:1'), covar=tensor([0.0156, 0.0288, 0.0370, 0.0208, 0.0184, 0.0231, 0.0223, 0.0301], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0037, 0.0036, 0.0038, 0.0037, 0.0034, 0.0038, 0.0046], + device='cuda:1'), out_proj_covar=tensor([4.3775e-05, 4.1310e-05, 4.0472e-05, 4.1961e-05, 4.0600e-05, 3.8344e-05, + 4.2650e-05, 5.1008e-05], device='cuda:1') +2023-03-21 09:49:59,465 INFO [train.py:901] (1/2) Epoch 40, batch 1600, loss[loss=0.1419, simple_loss=0.2286, pruned_loss=0.0276, over 7283.00 frames. ], tot_loss[loss=0.1314, simple_loss=0.213, pruned_loss=0.02496, over 1441355.30 frames. ], batch size: 86, lr: 4.12e-03, grad_scale: 8.0 +2023-03-21 09:50:02,520 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 09:50:03,020 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 09:50:06,053 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 09:50:06,511 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.192e+02 1.612e+02 1.846e+02 2.181e+02 5.378e+02, threshold=3.692e+02, percent-clipped=1.0 +2023-03-21 09:50:15,116 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 09:50:19,127 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 09:50:25,975 INFO [train.py:901] (1/2) Epoch 40, batch 1650, loss[loss=0.1484, simple_loss=0.2188, pruned_loss=0.03904, over 7274.00 frames. ], tot_loss[loss=0.1308, simple_loss=0.2125, pruned_loss=0.02457, over 1443817.01 frames. ], batch size: 47, lr: 4.12e-03, grad_scale: 8.0 +2023-03-21 09:50:28,125 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111791.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 09:50:28,489 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 09:50:34,176 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-03-21 09:50:44,735 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111824.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:50:45,638 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 09:50:49,594 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 09:50:51,071 INFO [train.py:901] (1/2) Epoch 40, batch 1700, loss[loss=0.1419, simple_loss=0.2261, pruned_loss=0.02886, over 7354.00 frames. ], tot_loss[loss=0.1307, simple_loss=0.2124, pruned_loss=0.02451, over 1443064.38 frames. ], batch size: 63, lr: 4.12e-03, grad_scale: 8.0 +2023-03-21 09:50:58,016 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.630e+02 2.034e+02 2.442e+02 3.825e+02, threshold=4.068e+02, percent-clipped=2.0 +2023-03-21 09:50:58,632 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111852.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 09:50:59,973 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 09:51:04,103 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6449, 2.4024, 2.4843, 3.7890, 1.9812, 3.5819, 1.3785, 3.3871], + device='cuda:1'), covar=tensor([0.0215, 0.1501, 0.1731, 0.0177, 0.3689, 0.0249, 0.1306, 0.0375], + device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0243, 0.0255, 0.0204, 0.0247, 0.0214, 0.0223, 0.0223], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:51:14,137 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111881.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:51:17,114 INFO [train.py:901] (1/2) Epoch 40, batch 1750, loss[loss=0.1396, simple_loss=0.2179, pruned_loss=0.0306, over 7278.00 frames. ], tot_loss[loss=0.1301, simple_loss=0.2114, pruned_loss=0.02433, over 1442719.45 frames. ], batch size: 57, lr: 4.12e-03, grad_scale: 16.0 +2023-03-21 09:51:25,205 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 09:51:26,228 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 09:51:32,511 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1529, 2.4623, 1.8808, 2.6108, 2.3591, 2.4389, 2.0938, 2.4052], + device='cuda:1'), covar=tensor([0.2100, 0.1151, 0.3883, 0.0718, 0.0316, 0.0257, 0.0336, 0.0390], + device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0229, 0.0244, 0.0257, 0.0198, 0.0198, 0.0217, 0.0223], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:51:38,279 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=111929.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:51:42,240 INFO [train.py:901] (1/2) Epoch 40, batch 1800, loss[loss=0.1364, simple_loss=0.2134, pruned_loss=0.02969, over 7323.00 frames. ], tot_loss[loss=0.1297, simple_loss=0.2109, pruned_loss=0.02426, over 1441149.43 frames. ], batch size: 49, lr: 4.12e-03, grad_scale: 16.0 +2023-03-21 09:51:47,683 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 09:51:49,138 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.360e+02 1.770e+02 2.065e+02 2.371e+02 4.234e+02, threshold=4.130e+02, percent-clipped=1.0 +2023-03-21 09:51:55,765 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4662, 2.3023, 2.3474, 3.5413, 1.8938, 3.4032, 1.3774, 3.0388], + device='cuda:1'), covar=tensor([0.0200, 0.1422, 0.1804, 0.0241, 0.3864, 0.0340, 0.1357, 0.0419], + device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0245, 0.0257, 0.0207, 0.0250, 0.0215, 0.0225, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:52:02,333 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 09:52:08,226 INFO [train.py:901] (1/2) Epoch 40, batch 1850, loss[loss=0.1397, simple_loss=0.2254, pruned_loss=0.02703, over 7227.00 frames. ], tot_loss[loss=0.1294, simple_loss=0.2106, pruned_loss=0.0241, over 1439912.50 frames. ], batch size: 93, lr: 4.12e-03, grad_scale: 8.0 +2023-03-21 09:52:11,331 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1285, 2.6910, 3.1114, 2.9822, 3.2369, 3.0079, 2.7478, 3.2449], + device='cuda:1'), covar=tensor([0.1340, 0.0863, 0.1373, 0.1393, 0.0926, 0.1132, 0.1819, 0.1038], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0068, 0.0052, 0.0052, 0.0050, 0.0050, 0.0069, 0.0051], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:1') +2023-03-21 09:52:12,720 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 09:52:30,533 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 +2023-03-21 09:52:33,218 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 09:52:37,750 INFO [train.py:901] (1/2) Epoch 40, batch 1900, loss[loss=0.1349, simple_loss=0.2189, pruned_loss=0.02544, over 7282.00 frames. ], tot_loss[loss=0.1302, simple_loss=0.2113, pruned_loss=0.02457, over 1439959.70 frames. ], batch size: 68, lr: 4.12e-03, grad_scale: 8.0 +2023-03-21 09:52:45,232 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.221e+02 1.755e+02 2.099e+02 2.615e+02 5.453e+02, threshold=4.198e+02, percent-clipped=1.0 +2023-03-21 09:52:46,628 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 +2023-03-21 09:52:52,975 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112066.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:52:57,908 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 09:52:58,521 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7512, 3.1122, 2.7633, 2.8626, 2.8687, 2.5186, 2.9209, 2.7028], + device='cuda:1'), covar=tensor([0.0771, 0.0423, 0.0932, 0.0873, 0.0845, 0.0827, 0.1215, 0.1274], + device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0060, 0.0069, 0.0061, 0.0058, 0.0063, 0.0057, 0.0055], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:53:03,467 INFO [train.py:901] (1/2) Epoch 40, batch 1950, loss[loss=0.1299, simple_loss=0.2071, pruned_loss=0.02639, over 7373.00 frames. ], tot_loss[loss=0.1299, simple_loss=0.211, pruned_loss=0.02441, over 1439184.20 frames. ], batch size: 51, lr: 4.12e-03, grad_scale: 8.0 +2023-03-21 09:53:09,332 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 09:53:13,850 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 09:53:14,807 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 09:53:21,974 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112124.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:53:23,490 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112127.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:53:28,389 INFO [train.py:901] (1/2) Epoch 40, batch 2000, loss[loss=0.1232, simple_loss=0.1976, pruned_loss=0.0244, over 7231.00 frames. ], tot_loss[loss=0.1304, simple_loss=0.2115, pruned_loss=0.02468, over 1440385.92 frames. ], batch size: 45, lr: 4.11e-03, grad_scale: 8.0 +2023-03-21 09:53:30,854 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 09:53:34,566 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112147.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 09:53:36,959 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.013e+02 1.664e+02 1.933e+02 2.332e+02 4.499e+02, threshold=3.866e+02, percent-clipped=2.0 +2023-03-21 09:53:43,078 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 09:53:47,172 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=112172.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:53:51,167 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 09:53:54,638 INFO [train.py:901] (1/2) Epoch 40, batch 2050, loss[loss=0.1232, simple_loss=0.2068, pruned_loss=0.01984, over 7253.00 frames. ], tot_loss[loss=0.1302, simple_loss=0.2113, pruned_loss=0.02457, over 1442217.61 frames. ], batch size: 47, lr: 4.11e-03, grad_scale: 8.0 +2023-03-21 09:53:55,858 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2912, 3.0875, 3.2061, 3.3235, 2.8762, 2.9161, 3.4550, 2.4203], + device='cuda:1'), covar=tensor([0.0541, 0.0608, 0.0810, 0.0773, 0.0727, 0.1101, 0.0835, 0.2462], + device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0336, 0.0273, 0.0355, 0.0286, 0.0286, 0.0347, 0.0244], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:54:21,551 INFO [train.py:901] (1/2) Epoch 40, batch 2100, loss[loss=0.126, simple_loss=0.2048, pruned_loss=0.02359, over 7249.00 frames. ], tot_loss[loss=0.1303, simple_loss=0.2113, pruned_loss=0.02464, over 1441199.12 frames. ], batch size: 45, lr: 4.11e-03, grad_scale: 8.0 +2023-03-21 09:54:24,984 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 09:54:27,983 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 09:54:28,982 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.301e+02 1.776e+02 2.099e+02 2.556e+02 5.854e+02, threshold=4.198e+02, percent-clipped=3.0 +2023-03-21 09:54:37,129 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9411, 2.1903, 1.8127, 2.4900, 2.5336, 2.4586, 2.4517, 2.4720], + device='cuda:1'), covar=tensor([0.2104, 0.1137, 0.3862, 0.0914, 0.0410, 0.0396, 0.0546, 0.0468], + device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0233, 0.0248, 0.0260, 0.0201, 0.0202, 0.0222, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 09:54:46,386 INFO [train.py:901] (1/2) Epoch 40, batch 2150, loss[loss=0.1462, simple_loss=0.2279, pruned_loss=0.03228, over 7335.00 frames. ], tot_loss[loss=0.1311, simple_loss=0.2124, pruned_loss=0.02492, over 1442029.01 frames. ], batch size: 61, lr: 4.11e-03, grad_scale: 8.0 +2023-03-21 09:55:12,681 INFO [train.py:901] (1/2) Epoch 40, batch 2200, loss[loss=0.1327, simple_loss=0.2135, pruned_loss=0.02599, over 7283.00 frames. ], tot_loss[loss=0.1311, simple_loss=0.2122, pruned_loss=0.02504, over 1443860.75 frames. ], batch size: 77, lr: 4.11e-03, grad_scale: 8.0 +2023-03-21 09:55:13,212 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 09:55:20,090 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.190e+02 1.737e+02 2.025e+02 2.418e+02 3.757e+02, threshold=4.050e+02, percent-clipped=0.0 +2023-03-21 09:55:27,214 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0988, 3.1621, 3.2975, 3.2269, 3.4335, 3.2491, 2.8339, 3.2641], + device='cuda:1'), covar=tensor([0.1614, 0.0765, 0.1324, 0.1657, 0.0900, 0.1011, 0.2195, 0.1657], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0069, 0.0052, 0.0052, 0.0051, 0.0050, 0.0070, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 09:55:36,809 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.1140, 4.6284, 4.3718, 5.0818, 4.8411, 4.9706, 4.2832, 4.6816], + device='cuda:1'), covar=tensor([0.0810, 0.2380, 0.2529, 0.0955, 0.0955, 0.1082, 0.0824, 0.1122], + device='cuda:1'), in_proj_covar=tensor([0.0260, 0.0391, 0.0293, 0.0303, 0.0227, 0.0366, 0.0229, 0.0274], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:55:37,773 INFO [train.py:901] (1/2) Epoch 40, batch 2250, loss[loss=0.1274, simple_loss=0.2108, pruned_loss=0.02198, over 7328.00 frames. ], tot_loss[loss=0.1318, simple_loss=0.2129, pruned_loss=0.02535, over 1442528.44 frames. ], batch size: 59, lr: 4.11e-03, grad_scale: 8.0 +2023-03-21 09:55:38,956 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112389.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:55:39,541 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-03-21 09:55:45,803 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 09:55:46,898 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 09:55:57,005 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112422.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:55:58,942 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 09:56:04,566 INFO [train.py:901] (1/2) Epoch 40, batch 2300, loss[loss=0.1264, simple_loss=0.2083, pruned_loss=0.0223, over 7311.00 frames. ], tot_loss[loss=0.132, simple_loss=0.213, pruned_loss=0.02545, over 1443641.95 frames. ], batch size: 75, lr: 4.11e-03, grad_scale: 8.0 +2023-03-21 09:56:09,712 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112447.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 09:56:11,259 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112450.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:56:12,108 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.304e+02 1.811e+02 2.116e+02 2.493e+02 3.897e+02, threshold=4.232e+02, percent-clipped=0.0 +2023-03-21 09:56:29,675 INFO [train.py:901] (1/2) Epoch 40, batch 2350, loss[loss=0.1414, simple_loss=0.2253, pruned_loss=0.02873, over 7314.00 frames. ], tot_loss[loss=0.1316, simple_loss=0.2128, pruned_loss=0.02516, over 1442547.75 frames. ], batch size: 59, lr: 4.11e-03, grad_scale: 8.0 +2023-03-21 09:56:34,416 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=112495.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 09:56:45,987 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 09:56:52,551 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 09:56:55,519 INFO [train.py:901] (1/2) Epoch 40, batch 2400, loss[loss=0.1141, simple_loss=0.1945, pruned_loss=0.01689, over 7147.00 frames. ], tot_loss[loss=0.1308, simple_loss=0.212, pruned_loss=0.02481, over 1441485.44 frames. ], batch size: 41, lr: 4.11e-03, grad_scale: 8.0 +2023-03-21 09:57:02,551 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 09:57:03,016 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.709e+02 1.972e+02 2.399e+02 3.691e+02, threshold=3.945e+02, percent-clipped=0.0 +2023-03-21 09:57:03,114 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.2609, 4.7192, 4.5107, 5.1666, 4.9882, 5.1111, 4.3974, 4.8273], + device='cuda:1'), covar=tensor([0.0706, 0.2580, 0.2430, 0.1059, 0.0995, 0.1180, 0.0902, 0.0981], + device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0396, 0.0295, 0.0308, 0.0229, 0.0370, 0.0232, 0.0276], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:57:05,533 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 09:57:08,155 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112562.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:57:15,523 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112577.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:57:19,242 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4908, 5.0295, 5.0990, 5.0824, 4.8984, 4.5340, 5.1183, 4.9362], + device='cuda:1'), covar=tensor([0.0501, 0.0401, 0.0399, 0.0424, 0.0345, 0.0399, 0.0327, 0.0463], + device='cuda:1'), in_proj_covar=tensor([0.0220, 0.0260, 0.0203, 0.0204, 0.0161, 0.0230, 0.0213, 0.0150], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:57:21,721 INFO [train.py:901] (1/2) Epoch 40, batch 2450, loss[loss=0.1353, simple_loss=0.2176, pruned_loss=0.02652, over 7315.00 frames. ], tot_loss[loss=0.1307, simple_loss=0.2119, pruned_loss=0.02476, over 1442109.66 frames. ], batch size: 59, lr: 4.11e-03, grad_scale: 8.0 +2023-03-21 09:57:24,421 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8818, 1.7765, 2.4022, 2.5084, 2.3384, 2.3740, 2.2540, 2.4403], + device='cuda:1'), covar=tensor([0.4895, 0.2498, 0.1680, 0.2266, 0.3427, 0.3136, 0.2358, 0.3708], + device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0080, 0.0072, 0.0065, 0.0064, 0.0064, 0.0105, 0.0069], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 09:57:34,626 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 09:57:37,280 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112617.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:57:40,367 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112623.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:57:47,170 INFO [train.py:901] (1/2) Epoch 40, batch 2500, loss[loss=0.1235, simple_loss=0.2079, pruned_loss=0.01956, over 7234.00 frames. ], tot_loss[loss=0.1301, simple_loss=0.2116, pruned_loss=0.02429, over 1441404.60 frames. ], batch size: 89, lr: 4.11e-03, grad_scale: 8.0 +2023-03-21 09:57:47,841 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112638.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 09:57:54,735 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.154e+02 1.621e+02 1.878e+02 2.077e+02 3.587e+02, threshold=3.756e+02, percent-clipped=0.0 +2023-03-21 09:57:58,176 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-03-21 09:57:59,298 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 09:58:09,194 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112678.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 09:58:13,471 INFO [train.py:901] (1/2) Epoch 40, batch 2550, loss[loss=0.1537, simple_loss=0.2256, pruned_loss=0.0409, over 7335.00 frames. ], tot_loss[loss=0.1303, simple_loss=0.2114, pruned_loss=0.02456, over 1438707.75 frames. ], batch size: 54, lr: 4.10e-03, grad_scale: 8.0 +2023-03-21 09:58:22,550 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9919, 4.1945, 3.9234, 4.2251, 3.8025, 4.2856, 4.5463, 4.5732], + device='cuda:1'), covar=tensor([0.0223, 0.0145, 0.0230, 0.0137, 0.0441, 0.0197, 0.0201, 0.0160], + device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0129, 0.0122, 0.0125, 0.0115, 0.0103, 0.0100, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:58:26,124 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6891, 2.1448, 1.8599, 1.9589, 2.1548, 1.7293, 1.8224, 1.4812], + device='cuda:1'), covar=tensor([0.0250, 0.0228, 0.0286, 0.0300, 0.0211, 0.0198, 0.0246, 0.0285], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0037, 0.0036, 0.0038, 0.0037, 0.0034, 0.0038, 0.0046], + device='cuda:1'), out_proj_covar=tensor([4.3472e-05, 4.1062e-05, 4.0529e-05, 4.1552e-05, 4.0532e-05, 3.8224e-05, + 4.2452e-05, 5.0574e-05], device='cuda:1') +2023-03-21 09:58:31,052 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112722.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:58:38,323 INFO [train.py:901] (1/2) Epoch 40, batch 2600, loss[loss=0.1402, simple_loss=0.2163, pruned_loss=0.03203, over 7271.00 frames. ], tot_loss[loss=0.1298, simple_loss=0.211, pruned_loss=0.02427, over 1439752.74 frames. ], batch size: 47, lr: 4.10e-03, grad_scale: 8.0 +2023-03-21 09:58:42,285 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112745.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:58:45,635 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.606e+02 1.959e+02 2.340e+02 7.874e+02, threshold=3.917e+02, percent-clipped=2.0 +2023-03-21 09:58:54,537 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=112770.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 09:59:02,870 INFO [train.py:901] (1/2) Epoch 40, batch 2650, loss[loss=0.155, simple_loss=0.2383, pruned_loss=0.03582, over 7159.00 frames. ], tot_loss[loss=0.1294, simple_loss=0.2107, pruned_loss=0.02409, over 1437883.91 frames. ], batch size: 98, lr: 4.10e-03, grad_scale: 8.0 +2023-03-21 09:59:18,076 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2329, 2.2565, 2.6170, 2.1108, 2.4554, 2.1967, 2.1173, 1.8721], + device='cuda:1'), covar=tensor([0.0503, 0.0435, 0.0220, 0.0385, 0.0358, 0.0426, 0.0335, 0.0292], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0039, 0.0040, 0.0039, 0.0036, 0.0037, 0.0042, 0.0042], + device='cuda:1'), out_proj_covar=tensor([1.0298e-04, 1.0077e-04, 9.9785e-05, 9.8338e-05, 9.6156e-05, 9.6257e-05, + 1.0471e-04, 1.0528e-04], device='cuda:1') +2023-03-21 09:59:27,047 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9932, 3.6927, 3.6878, 3.6812, 3.6894, 3.4948, 3.8724, 3.5000], + device='cuda:1'), covar=tensor([0.0151, 0.0186, 0.0125, 0.0204, 0.0431, 0.0132, 0.0159, 0.0176], + device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0108, 0.0108, 0.0095, 0.0186, 0.0113, 0.0109, 0.0119], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 09:59:28,799 INFO [train.py:901] (1/2) Epoch 40, batch 2700, loss[loss=0.1187, simple_loss=0.1973, pruned_loss=0.02008, over 7321.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.2111, pruned_loss=0.02404, over 1441421.65 frames. ], batch size: 44, lr: 4.10e-03, grad_scale: 8.0 +2023-03-21 09:59:36,113 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 1.707e+02 2.016e+02 2.345e+02 3.632e+02, threshold=4.032e+02, percent-clipped=0.0 +2023-03-21 09:59:53,424 INFO [train.py:901] (1/2) Epoch 40, batch 2750, loss[loss=0.1273, simple_loss=0.2081, pruned_loss=0.02324, over 7297.00 frames. ], tot_loss[loss=0.1298, simple_loss=0.2114, pruned_loss=0.02409, over 1441872.52 frames. ], batch size: 66, lr: 4.10e-03, grad_scale: 8.0 +2023-03-21 10:00:08,765 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112918.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:00:10,297 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112921.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:00:14,949 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.04 vs. limit=5.0 +2023-03-21 10:00:16,116 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112933.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 10:00:17,997 INFO [train.py:901] (1/2) Epoch 40, batch 2800, loss[loss=0.1068, simple_loss=0.1785, pruned_loss=0.01757, over 6997.00 frames. ], tot_loss[loss=0.1297, simple_loss=0.2111, pruned_loss=0.02414, over 1443578.08 frames. ], batch size: 35, lr: 4.10e-03, grad_scale: 8.0 +2023-03-21 10:00:19,304 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 +2023-03-21 10:00:25,334 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.722e+02 1.974e+02 2.339e+02 4.614e+02, threshold=3.948e+02, percent-clipped=1.0 +2023-03-21 10:00:40,640 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 10:00:40,691 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 10:00:40,716 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 10:00:40,732 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 10:00:40,936 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 10:00:40,952 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 10:00:40,962 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 10:00:41,061 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 10:00:41,133 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 10:00:41,148 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 10:00:41,474 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 10:00:41,631 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 10:00:41,638 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 10:00:41,645 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 10:00:41,788 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 10:00:41,791 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 10:00:41,852 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 10:00:49,023 INFO [train.py:901] (1/2) Epoch 41, batch 0, loss[loss=0.1424, simple_loss=0.2214, pruned_loss=0.03172, over 7249.00 frames. ], tot_loss[loss=0.1424, simple_loss=0.2214, pruned_loss=0.03172, over 7249.00 frames. ], batch size: 55, lr: 4.05e-03, grad_scale: 8.0 +2023-03-21 10:00:49,023 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 10:00:56,388 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6659, 3.7113, 2.8827, 3.4403, 2.8412, 2.4676, 1.9327, 3.7416], + device='cuda:1'), covar=tensor([0.0061, 0.0060, 0.0170, 0.0073, 0.0175, 0.0574, 0.0760, 0.0056], + device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0096, 0.0113, 0.0094, 0.0131, 0.0135, 0.0129, 0.0107], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 10:01:05,572 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7591, 3.9195, 3.6831, 3.9736, 3.6463, 3.9411, 4.1715, 4.2210], + device='cuda:1'), covar=tensor([0.0198, 0.0177, 0.0245, 0.0147, 0.0305, 0.0205, 0.0212, 0.0145], + device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0129, 0.0123, 0.0125, 0.0115, 0.0102, 0.0100, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 10:01:12,852 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3865, 4.0258, 4.0593, 4.0540, 4.0936, 3.9037, 4.2148, 3.8867], + device='cuda:1'), covar=tensor([0.0120, 0.0163, 0.0117, 0.0178, 0.0457, 0.0131, 0.0151, 0.0174], + device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0107, 0.0108, 0.0095, 0.0186, 0.0113, 0.0109, 0.0119], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 10:01:15,063 INFO [train.py:935] (1/2) Epoch 41, validation: loss=0.1652, simple_loss=0.2574, pruned_loss=0.03654, over 1622729.00 frames. +2023-03-21 10:01:15,064 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 10:01:21,210 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112973.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 10:01:22,130 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 10:01:22,264 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5434, 1.4606, 1.6288, 1.9453, 1.8035, 1.9076, 1.4245, 1.9129], + device='cuda:1'), covar=tensor([0.2530, 0.3678, 0.1435, 0.1729, 0.1697, 0.1840, 0.2638, 0.2084], + device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0079, 0.0071, 0.0063, 0.0064, 0.0063, 0.0102, 0.0068], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:01:25,878 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112982.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:01:32,682 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 10:01:40,018 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 10:01:40,507 INFO [train.py:901] (1/2) Epoch 41, batch 50, loss[loss=0.1305, simple_loss=0.2148, pruned_loss=0.02308, over 7354.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2102, pruned_loss=0.02329, over 325644.10 frames. ], batch size: 73, lr: 4.05e-03, grad_scale: 8.0 +2023-03-21 10:01:42,094 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 10:01:44,554 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 10:01:55,850 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3552, 3.9234, 3.9467, 4.0368, 4.0224, 3.8766, 4.1946, 3.7197], + device='cuda:1'), covar=tensor([0.0184, 0.0189, 0.0139, 0.0176, 0.0440, 0.0127, 0.0160, 0.0207], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0106, 0.0107, 0.0094, 0.0183, 0.0111, 0.0108, 0.0118], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 10:01:58,874 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113045.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:02:02,324 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.308e+02 1.789e+02 2.061e+02 2.499e+02 5.277e+02, threshold=4.122e+02, percent-clipped=1.0 +2023-03-21 10:02:02,890 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 10:02:03,352 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 10:02:06,877 INFO [train.py:901] (1/2) Epoch 41, batch 100, loss[loss=0.1172, simple_loss=0.1987, pruned_loss=0.01786, over 7303.00 frames. ], tot_loss[loss=0.1283, simple_loss=0.2092, pruned_loss=0.0237, over 572219.22 frames. ], batch size: 49, lr: 4.05e-03, grad_scale: 8.0 +2023-03-21 10:02:22,752 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=113093.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:02:23,066 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 +2023-03-21 10:02:32,458 INFO [train.py:901] (1/2) Epoch 41, batch 150, loss[loss=0.138, simple_loss=0.2181, pruned_loss=0.02892, over 7295.00 frames. ], tot_loss[loss=0.1303, simple_loss=0.2115, pruned_loss=0.02458, over 765101.03 frames. ], batch size: 66, lr: 4.05e-03, grad_scale: 8.0 +2023-03-21 10:02:45,898 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-03-21 10:02:53,730 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.262e+02 1.658e+02 2.006e+02 2.352e+02 4.117e+02, threshold=4.012e+02, percent-clipped=0.0 +2023-03-21 10:02:58,226 INFO [train.py:901] (1/2) Epoch 41, batch 200, loss[loss=0.1252, simple_loss=0.1998, pruned_loss=0.02529, over 7253.00 frames. ], tot_loss[loss=0.1307, simple_loss=0.2123, pruned_loss=0.02459, over 915647.13 frames. ], batch size: 47, lr: 4.05e-03, grad_scale: 8.0 +2023-03-21 10:03:02,206 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 10:03:06,769 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 10:03:13,292 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 10:03:24,892 INFO [train.py:901] (1/2) Epoch 41, batch 250, loss[loss=0.1304, simple_loss=0.2124, pruned_loss=0.02422, over 7237.00 frames. ], tot_loss[loss=0.1294, simple_loss=0.211, pruned_loss=0.02391, over 1031311.88 frames. ], batch size: 93, lr: 4.05e-03, grad_scale: 8.0 +2023-03-21 10:03:26,896 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 10:03:28,453 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113218.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:03:36,070 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113233.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:03:36,581 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113234.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:03:42,965 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 +2023-03-21 10:03:43,765 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1638, 2.8103, 3.2564, 3.2609, 3.3935, 3.1070, 2.8056, 3.2513], + device='cuda:1'), covar=tensor([0.1396, 0.0767, 0.1029, 0.1101, 0.0648, 0.0764, 0.2228, 0.1431], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0070, 0.0053, 0.0052, 0.0052, 0.0051, 0.0071, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 10:03:46,067 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.202e+02 1.702e+02 2.019e+02 2.239e+02 3.429e+02, threshold=4.038e+02, percent-clipped=0.0 +2023-03-21 10:03:47,662 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 10:03:50,101 INFO [train.py:901] (1/2) Epoch 41, batch 300, loss[loss=0.1497, simple_loss=0.2245, pruned_loss=0.03746, over 7362.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2094, pruned_loss=0.02367, over 1118241.78 frames. ], batch size: 63, lr: 4.04e-03, grad_scale: 4.0 +2023-03-21 10:03:52,735 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=113266.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:03:56,338 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113273.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 10:03:56,746 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 10:03:58,972 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113277.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:04:00,947 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=113281.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:04:08,753 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113295.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:04:16,638 INFO [train.py:901] (1/2) Epoch 41, batch 350, loss[loss=0.1406, simple_loss=0.2193, pruned_loss=0.0309, over 7334.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2094, pruned_loss=0.02353, over 1192471.75 frames. ], batch size: 54, lr: 4.04e-03, grad_scale: 4.0 +2023-03-21 10:04:21,733 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=113321.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:04:23,488 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-03-21 10:04:31,715 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 10:04:34,910 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8833, 3.7872, 2.8973, 3.4832, 2.6272, 2.2022, 1.9101, 3.8528], + device='cuda:1'), covar=tensor([0.0050, 0.0058, 0.0187, 0.0070, 0.0230, 0.0557, 0.0614, 0.0056], + device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0096, 0.0114, 0.0095, 0.0132, 0.0137, 0.0130, 0.0107], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 10:04:35,949 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6580, 2.0087, 1.6560, 1.9669, 2.1332, 1.7295, 1.7598, 1.4485], + device='cuda:1'), covar=tensor([0.0187, 0.0219, 0.0388, 0.0197, 0.0128, 0.0147, 0.0246, 0.0201], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0037, 0.0037, 0.0038, 0.0037, 0.0035, 0.0038, 0.0046], + device='cuda:1'), out_proj_covar=tensor([4.3696e-05, 4.1553e-05, 4.1276e-05, 4.2210e-05, 4.0565e-05, 3.8485e-05, + 4.2900e-05, 5.0998e-05], device='cuda:1') +2023-03-21 10:04:37,770 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.063e+02 1.665e+02 1.929e+02 2.230e+02 3.805e+02, threshold=3.858e+02, percent-clipped=0.0 +2023-03-21 10:04:38,067 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-03-21 10:04:40,081 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-03-21 10:04:41,798 INFO [train.py:901] (1/2) Epoch 41, batch 400, loss[loss=0.1071, simple_loss=0.1776, pruned_loss=0.01824, over 7046.00 frames. ], tot_loss[loss=0.1287, simple_loss=0.2099, pruned_loss=0.02377, over 1248244.55 frames. ], batch size: 35, lr: 4.04e-03, grad_scale: 8.0 +2023-03-21 10:04:58,863 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1510, 3.6701, 4.1043, 4.4219, 4.2692, 4.2411, 4.3773, 4.0726], + device='cuda:1'), covar=tensor([0.0041, 0.0132, 0.0042, 0.0034, 0.0035, 0.0043, 0.0033, 0.0063], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0071, 0.0059, 0.0057, 0.0055, 0.0061, 0.0048, 0.0078], + device='cuda:1'), out_proj_covar=tensor([8.2063e-05, 1.4361e-04, 1.0550e-04, 9.7631e-05, 9.3926e-05, 1.0616e-04, + 9.0669e-05, 1.4413e-04], device='cuda:1') +2023-03-21 10:05:08,139 INFO [train.py:901] (1/2) Epoch 41, batch 450, loss[loss=0.1413, simple_loss=0.2309, pruned_loss=0.02585, over 6672.00 frames. ], tot_loss[loss=0.1287, simple_loss=0.2101, pruned_loss=0.02369, over 1292101.03 frames. ], batch size: 106, lr: 4.04e-03, grad_scale: 8.0 +2023-03-21 10:05:13,142 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 10:05:13,701 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 10:05:19,037 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.97 vs. limit=5.0 +2023-03-21 10:05:19,067 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-03-21 10:05:29,865 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.168e+02 1.845e+02 2.130e+02 2.461e+02 6.042e+02, threshold=4.260e+02, percent-clipped=2.0 +2023-03-21 10:05:33,951 INFO [train.py:901] (1/2) Epoch 41, batch 500, loss[loss=0.1338, simple_loss=0.218, pruned_loss=0.0248, over 7350.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2092, pruned_loss=0.02354, over 1322586.19 frames. ], batch size: 63, lr: 4.04e-03, grad_scale: 8.0 +2023-03-21 10:05:44,647 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 +2023-03-21 10:05:46,446 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4824, 4.9907, 5.0720, 5.0098, 4.8897, 4.5179, 5.0801, 4.9058], + device='cuda:1'), covar=tensor([0.0467, 0.0416, 0.0421, 0.0512, 0.0360, 0.0412, 0.0385, 0.0426], + device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0266, 0.0206, 0.0207, 0.0164, 0.0233, 0.0217, 0.0151], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:05:48,387 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 10:05:49,919 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 10:05:50,442 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 10:05:52,528 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 10:05:54,360 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 +2023-03-21 10:05:57,023 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 10:06:00,023 INFO [train.py:901] (1/2) Epoch 41, batch 550, loss[loss=0.1182, simple_loss=0.1915, pruned_loss=0.02243, over 7133.00 frames. ], tot_loss[loss=0.129, simple_loss=0.2101, pruned_loss=0.02391, over 1351017.91 frames. ], batch size: 41, lr: 4.04e-03, grad_scale: 8.0 +2023-03-21 10:06:09,564 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 10:06:18,369 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 10:06:22,450 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.173e+02 1.584e+02 1.802e+02 2.149e+02 4.424e+02, threshold=3.603e+02, percent-clipped=1.0 +2023-03-21 10:06:22,471 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 10:06:26,486 INFO [train.py:901] (1/2) Epoch 41, batch 600, loss[loss=0.1213, simple_loss=0.2098, pruned_loss=0.0164, over 7321.00 frames. ], tot_loss[loss=0.1295, simple_loss=0.2108, pruned_loss=0.02415, over 1373527.93 frames. ], batch size: 80, lr: 4.04e-03, grad_scale: 8.0 +2023-03-21 10:06:29,004 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 10:06:34,722 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113577.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:06:41,258 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113590.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:06:44,184 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 10:06:52,130 INFO [train.py:901] (1/2) Epoch 41, batch 650, loss[loss=0.1293, simple_loss=0.2176, pruned_loss=0.02053, over 7356.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.2108, pruned_loss=0.02415, over 1388085.91 frames. ], batch size: 73, lr: 4.04e-03, grad_scale: 8.0 +2023-03-21 10:06:52,680 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 10:06:55,245 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113617.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:06:59,262 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=113625.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:07:11,046 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 10:07:11,150 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113646.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:07:14,446 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.694e+02 1.988e+02 2.330e+02 3.626e+02, threshold=3.977e+02, percent-clipped=1.0 +2023-03-21 10:07:18,570 INFO [train.py:901] (1/2) Epoch 41, batch 700, loss[loss=0.1194, simple_loss=0.2056, pruned_loss=0.01663, over 7252.00 frames. ], tot_loss[loss=0.1299, simple_loss=0.2109, pruned_loss=0.02449, over 1398016.03 frames. ], batch size: 55, lr: 4.04e-03, grad_scale: 8.0 +2023-03-21 10:07:19,562 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 10:07:27,127 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113678.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:07:30,825 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-03-21 10:07:38,680 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4241, 3.4775, 2.4467, 3.7589, 2.8811, 3.4374, 1.7088, 2.5112], + device='cuda:1'), covar=tensor([0.0514, 0.0757, 0.2545, 0.0703, 0.0582, 0.0761, 0.4165, 0.1991], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0256, 0.0278, 0.0266, 0.0270, 0.0266, 0.0232, 0.0257], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 10:07:39,145 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3858, 4.1926, 3.9061, 3.9125, 3.6395, 2.4744, 2.0859, 4.3180], + device='cuda:1'), covar=tensor([0.0045, 0.0070, 0.0089, 0.0068, 0.0114, 0.0517, 0.0581, 0.0051], + device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0095, 0.0113, 0.0094, 0.0129, 0.0135, 0.0128, 0.0106], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 10:07:41,545 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 10:07:41,678 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113707.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:07:42,056 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 10:07:43,470 INFO [train.py:901] (1/2) Epoch 41, batch 750, loss[loss=0.1417, simple_loss=0.2171, pruned_loss=0.03311, over 7252.00 frames. ], tot_loss[loss=0.1306, simple_loss=0.2114, pruned_loss=0.0249, over 1409041.55 frames. ], batch size: 45, lr: 4.04e-03, grad_scale: 8.0 +2023-03-21 10:07:57,630 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 10:08:02,018 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 10:08:05,449 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.674e+02 1.983e+02 2.352e+02 7.977e+02, threshold=3.966e+02, percent-clipped=3.0 +2023-03-21 10:08:08,011 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 10:08:09,521 INFO [train.py:901] (1/2) Epoch 41, batch 800, loss[loss=0.1109, simple_loss=0.1848, pruned_loss=0.01852, over 7146.00 frames. ], tot_loss[loss=0.1306, simple_loss=0.2115, pruned_loss=0.02488, over 1417238.21 frames. ], batch size: 39, lr: 4.04e-03, grad_scale: 8.0 +2023-03-21 10:08:09,536 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 10:08:19,502 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 10:08:36,114 INFO [train.py:901] (1/2) Epoch 41, batch 850, loss[loss=0.1273, simple_loss=0.2085, pruned_loss=0.02308, over 7271.00 frames. ], tot_loss[loss=0.1309, simple_loss=0.2122, pruned_loss=0.02485, over 1424353.92 frames. ], batch size: 52, lr: 4.03e-03, grad_scale: 8.0 +2023-03-21 10:08:39,706 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 10:08:40,170 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 10:08:45,241 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 10:08:45,903 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8689, 2.7493, 2.9788, 2.7245, 2.5845, 2.6992, 2.7522, 2.5078], + device='cuda:1'), covar=tensor([0.0304, 0.0625, 0.0383, 0.0291, 0.1443, 0.0594, 0.0352, 0.0326], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0039, 0.0040, 0.0039, 0.0037, 0.0037, 0.0042, 0.0042], + device='cuda:1'), out_proj_covar=tensor([1.0350e-04, 1.0197e-04, 1.0148e-04, 9.8870e-05, 9.8030e-05, 9.7077e-05, + 1.0541e-04, 1.0692e-04], device='cuda:1') +2023-03-21 10:08:48,905 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 10:08:57,501 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.207e+02 1.665e+02 1.936e+02 2.284e+02 3.420e+02, threshold=3.871e+02, percent-clipped=0.0 +2023-03-21 10:09:01,408 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-03-21 10:09:01,602 INFO [train.py:901] (1/2) Epoch 41, batch 900, loss[loss=0.12, simple_loss=0.2016, pruned_loss=0.01924, over 7343.00 frames. ], tot_loss[loss=0.1299, simple_loss=0.211, pruned_loss=0.02438, over 1427309.94 frames. ], batch size: 73, lr: 4.03e-03, grad_scale: 8.0 +2023-03-21 10:09:13,827 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 +2023-03-21 10:09:15,594 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5640, 5.0925, 5.1997, 5.1445, 4.9785, 4.6267, 5.1916, 5.0265], + device='cuda:1'), covar=tensor([0.0418, 0.0359, 0.0345, 0.0437, 0.0304, 0.0359, 0.0309, 0.0403], + device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0265, 0.0206, 0.0205, 0.0162, 0.0232, 0.0216, 0.0150], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:09:17,167 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113890.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:09:27,283 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 10:09:28,301 INFO [train.py:901] (1/2) Epoch 41, batch 950, loss[loss=0.133, simple_loss=0.217, pruned_loss=0.02449, over 7283.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.211, pruned_loss=0.02409, over 1431906.52 frames. ], batch size: 77, lr: 4.03e-03, grad_scale: 8.0 +2023-03-21 10:09:42,044 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=113938.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:09:49,397 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.076e+02 1.644e+02 2.020e+02 2.352e+02 7.244e+02, threshold=4.039e+02, percent-clipped=3.0 +2023-03-21 10:09:51,449 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 10:09:53,448 INFO [train.py:901] (1/2) Epoch 41, batch 1000, loss[loss=0.1272, simple_loss=0.2133, pruned_loss=0.0205, over 7283.00 frames. ], tot_loss[loss=0.1295, simple_loss=0.211, pruned_loss=0.02399, over 1434868.92 frames. ], batch size: 57, lr: 4.03e-03, grad_scale: 8.0 +2023-03-21 10:09:58,622 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7523, 1.5239, 1.8602, 2.0668, 1.9417, 1.9355, 1.5194, 2.1133], + device='cuda:1'), covar=tensor([0.3080, 0.4213, 0.1242, 0.1110, 0.1121, 0.0997, 0.2542, 0.1084], + device='cuda:1'), in_proj_covar=tensor([0.0080, 0.0081, 0.0073, 0.0066, 0.0066, 0.0065, 0.0105, 0.0069], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:10:00,237 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113973.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:10:11,903 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 10:10:15,743 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114002.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:10:20,138 INFO [train.py:901] (1/2) Epoch 41, batch 1050, loss[loss=0.1336, simple_loss=0.2161, pruned_loss=0.02552, over 7298.00 frames. ], tot_loss[loss=0.1295, simple_loss=0.2111, pruned_loss=0.02394, over 1435583.54 frames. ], batch size: 66, lr: 4.03e-03, grad_scale: 8.0 +2023-03-21 10:10:33,477 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 10:10:37,439 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 10:10:40,997 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.136e+02 1.735e+02 2.009e+02 2.345e+02 4.363e+02, threshold=4.019e+02, percent-clipped=1.0 +2023-03-21 10:10:45,252 INFO [train.py:901] (1/2) Epoch 41, batch 1100, loss[loss=0.1216, simple_loss=0.1981, pruned_loss=0.02254, over 7333.00 frames. ], tot_loss[loss=0.1286, simple_loss=0.2101, pruned_loss=0.02358, over 1436656.42 frames. ], batch size: 61, lr: 4.03e-03, grad_scale: 8.0 +2023-03-21 10:10:46,784 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 +2023-03-21 10:11:01,587 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-21 10:11:07,815 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 10:11:07,828 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 10:11:11,747 INFO [train.py:901] (1/2) Epoch 41, batch 1150, loss[loss=0.1295, simple_loss=0.2037, pruned_loss=0.02761, over 7287.00 frames. ], tot_loss[loss=0.1294, simple_loss=0.2107, pruned_loss=0.02406, over 1437051.44 frames. ], batch size: 57, lr: 4.03e-03, grad_scale: 8.0 +2023-03-21 10:11:19,869 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114127.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:11:20,778 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 10:11:21,288 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 10:11:33,376 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.308e+02 1.735e+02 1.978e+02 2.469e+02 4.023e+02, threshold=3.955e+02, percent-clipped=1.0 +2023-03-21 10:11:38,050 INFO [train.py:901] (1/2) Epoch 41, batch 1200, loss[loss=0.1412, simple_loss=0.2246, pruned_loss=0.0289, over 7374.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.2109, pruned_loss=0.02414, over 1440453.03 frames. ], batch size: 65, lr: 4.03e-03, grad_scale: 8.0 +2023-03-21 10:11:51,772 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114188.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:11:54,136 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 10:12:03,598 INFO [train.py:901] (1/2) Epoch 41, batch 1250, loss[loss=0.1289, simple_loss=0.2198, pruned_loss=0.01903, over 7351.00 frames. ], tot_loss[loss=0.1292, simple_loss=0.2107, pruned_loss=0.02387, over 1439951.27 frames. ], batch size: 73, lr: 4.03e-03, grad_scale: 8.0 +2023-03-21 10:12:17,554 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 10:12:22,613 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 10:12:24,715 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 10:12:25,678 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.306e+02 1.779e+02 2.107e+02 2.486e+02 4.646e+02, threshold=4.214e+02, percent-clipped=2.0 +2023-03-21 10:12:29,811 INFO [train.py:901] (1/2) Epoch 41, batch 1300, loss[loss=0.1224, simple_loss=0.2062, pruned_loss=0.01932, over 7285.00 frames. ], tot_loss[loss=0.1291, simple_loss=0.2106, pruned_loss=0.02384, over 1439278.33 frames. ], batch size: 86, lr: 4.03e-03, grad_scale: 8.0 +2023-03-21 10:12:32,529 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7156, 2.2416, 2.5156, 3.8032, 1.9683, 3.5556, 1.4404, 3.4785], + device='cuda:1'), covar=tensor([0.0209, 0.1653, 0.1839, 0.0213, 0.3860, 0.0300, 0.1342, 0.0418], + device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0249, 0.0262, 0.0211, 0.0251, 0.0218, 0.0225, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 10:12:36,007 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114273.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:12:41,092 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8922, 3.1365, 2.7111, 3.1392, 2.9015, 2.7860, 2.9095, 2.6450], + device='cuda:1'), covar=tensor([0.0539, 0.0595, 0.0931, 0.0969, 0.1433, 0.0495, 0.1096, 0.1471], + device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0059, 0.0068, 0.0060, 0.0057, 0.0062, 0.0057, 0.0054], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 10:12:47,183 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4039, 1.6812, 1.4899, 1.5755, 1.6937, 1.6237, 1.4899, 1.3500], + device='cuda:1'), covar=tensor([0.0155, 0.0170, 0.0171, 0.0143, 0.0106, 0.0128, 0.0185, 0.0188], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0037, 0.0035, 0.0038, 0.0036, 0.0034, 0.0037, 0.0046], + device='cuda:1'), out_proj_covar=tensor([4.3067e-05, 4.1427e-05, 4.0107e-05, 4.1787e-05, 3.9606e-05, 3.7659e-05, + 4.1913e-05, 5.0690e-05], device='cuda:1') +2023-03-21 10:12:48,995 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 10:12:50,557 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114302.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:12:50,997 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 10:12:54,037 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 10:12:55,001 INFO [train.py:901] (1/2) Epoch 41, batch 1350, loss[loss=0.1334, simple_loss=0.222, pruned_loss=0.02234, over 7282.00 frames. ], tot_loss[loss=0.1291, simple_loss=0.2105, pruned_loss=0.02384, over 1439529.61 frames. ], batch size: 57, lr: 4.03e-03, grad_scale: 8.0 +2023-03-21 10:13:00,103 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=114321.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:13:04,108 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 10:13:13,015 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5035, 5.0527, 5.1585, 5.0682, 4.8862, 4.5675, 5.1306, 4.9056], + device='cuda:1'), covar=tensor([0.0475, 0.0383, 0.0342, 0.0436, 0.0332, 0.0387, 0.0316, 0.0433], + device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0265, 0.0205, 0.0204, 0.0161, 0.0231, 0.0215, 0.0150], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:13:15,958 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=114350.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:13:17,358 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.320e+02 1.723e+02 1.987e+02 2.358e+02 3.719e+02, threshold=3.975e+02, percent-clipped=0.0 +2023-03-21 10:13:17,733 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.07 vs. limit=2.0 +2023-03-21 10:13:18,274 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-21 10:13:21,351 INFO [train.py:901] (1/2) Epoch 41, batch 1400, loss[loss=0.1392, simple_loss=0.2214, pruned_loss=0.02848, over 7347.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2097, pruned_loss=0.02348, over 1438777.58 frames. ], batch size: 73, lr: 4.02e-03, grad_scale: 8.0 +2023-03-21 10:13:31,694 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 +2023-03-21 10:13:36,869 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 10:13:42,091 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-03-21 10:13:46,775 INFO [train.py:901] (1/2) Epoch 41, batch 1450, loss[loss=0.1456, simple_loss=0.226, pruned_loss=0.03257, over 7269.00 frames. ], tot_loss[loss=0.1288, simple_loss=0.21, pruned_loss=0.02374, over 1440392.63 frames. ], batch size: 70, lr: 4.02e-03, grad_scale: 8.0 +2023-03-21 10:13:59,302 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-21 10:14:02,616 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 10:14:07,458 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.58 vs. limit=5.0 +2023-03-21 10:14:09,120 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.348e+02 1.681e+02 1.902e+02 2.503e+02 5.939e+02, threshold=3.804e+02, percent-clipped=2.0 +2023-03-21 10:14:13,119 INFO [train.py:901] (1/2) Epoch 41, batch 1500, loss[loss=0.1207, simple_loss=0.208, pruned_loss=0.01666, over 7239.00 frames. ], tot_loss[loss=0.1288, simple_loss=0.2102, pruned_loss=0.02366, over 1443304.55 frames. ], batch size: 89, lr: 4.02e-03, grad_scale: 8.0 +2023-03-21 10:14:19,188 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 10:14:24,334 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114483.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:14:39,028 INFO [train.py:901] (1/2) Epoch 41, batch 1550, loss[loss=0.1234, simple_loss=0.2105, pruned_loss=0.01808, over 7282.00 frames. ], tot_loss[loss=0.1293, simple_loss=0.2107, pruned_loss=0.02396, over 1443423.52 frames. ], batch size: 86, lr: 4.02e-03, grad_scale: 8.0 +2023-03-21 10:14:43,742 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 10:14:44,036 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 10:15:00,786 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.328e+02 1.777e+02 2.044e+02 2.457e+02 4.625e+02, threshold=4.087e+02, percent-clipped=1.0 +2023-03-21 10:15:04,694 INFO [train.py:901] (1/2) Epoch 41, batch 1600, loss[loss=0.1589, simple_loss=0.2477, pruned_loss=0.03508, over 6719.00 frames. ], tot_loss[loss=0.1293, simple_loss=0.2108, pruned_loss=0.02392, over 1444840.47 frames. ], batch size: 106, lr: 4.02e-03, grad_scale: 8.0 +2023-03-21 10:15:13,995 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 10:15:14,148 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8959, 1.7754, 2.2400, 2.2938, 2.3784, 2.2078, 2.2714, 2.4736], + device='cuda:1'), covar=tensor([0.3995, 0.5928, 0.1505, 0.1241, 0.1729, 0.3369, 0.2181, 0.1422], + device='cuda:1'), in_proj_covar=tensor([0.0081, 0.0082, 0.0073, 0.0067, 0.0067, 0.0068, 0.0107, 0.0070], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:15:14,521 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 10:15:16,954 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 10:15:28,023 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 10:15:31,065 INFO [train.py:901] (1/2) Epoch 41, batch 1650, loss[loss=0.1062, simple_loss=0.1882, pruned_loss=0.01211, over 7131.00 frames. ], tot_loss[loss=0.1294, simple_loss=0.211, pruned_loss=0.02386, over 1445773.34 frames. ], batch size: 41, lr: 4.02e-03, grad_scale: 8.0 +2023-03-21 10:15:32,576 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 10:15:39,743 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8500, 3.3379, 2.5535, 3.7020, 3.6648, 3.5883, 3.5180, 3.4548], + device='cuda:1'), covar=tensor([0.1689, 0.0718, 0.3411, 0.0825, 0.0399, 0.0316, 0.0594, 0.0519], + device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0229, 0.0242, 0.0254, 0.0196, 0.0201, 0.0217, 0.0222], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 10:15:40,576 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 10:15:52,152 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.270e+02 1.753e+02 2.036e+02 2.309e+02 4.360e+02, threshold=4.073e+02, percent-clipped=1.0 +2023-03-21 10:15:56,200 INFO [train.py:901] (1/2) Epoch 41, batch 1700, loss[loss=0.1362, simple_loss=0.2207, pruned_loss=0.0259, over 7256.00 frames. ], tot_loss[loss=0.1286, simple_loss=0.2102, pruned_loss=0.02356, over 1442912.12 frames. ], batch size: 89, lr: 4.02e-03, grad_scale: 8.0 +2023-03-21 10:15:57,201 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 10:16:01,236 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 10:16:13,546 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 10:16:22,614 INFO [train.py:901] (1/2) Epoch 41, batch 1750, loss[loss=0.1328, simple_loss=0.2174, pruned_loss=0.0241, over 7252.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2097, pruned_loss=0.02336, over 1442675.28 frames. ], batch size: 64, lr: 4.02e-03, grad_scale: 8.0 +2023-03-21 10:16:25,340 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114716.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:16:37,413 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 10:16:38,403 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 10:16:43,880 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.738e+02 2.103e+02 2.462e+02 3.793e+02, threshold=4.205e+02, percent-clipped=0.0 +2023-03-21 10:16:47,911 INFO [train.py:901] (1/2) Epoch 41, batch 1800, loss[loss=0.1182, simple_loss=0.2041, pruned_loss=0.01618, over 7300.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2096, pruned_loss=0.02324, over 1441098.27 frames. ], batch size: 49, lr: 4.02e-03, grad_scale: 8.0 +2023-03-21 10:16:49,050 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1403, 2.7978, 3.3087, 2.9587, 3.2506, 3.0235, 2.8243, 3.2357], + device='cuda:1'), covar=tensor([0.1226, 0.0754, 0.0971, 0.1484, 0.0926, 0.0958, 0.1417, 0.1525], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0070, 0.0052, 0.0052, 0.0051, 0.0051, 0.0070, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 10:16:56,732 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114777.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:16:59,119 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-21 10:17:00,438 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114783.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:17:01,330 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 10:17:14,680 INFO [train.py:901] (1/2) Epoch 41, batch 1850, loss[loss=0.1169, simple_loss=0.2077, pruned_loss=0.01301, over 7212.00 frames. ], tot_loss[loss=0.1288, simple_loss=0.2103, pruned_loss=0.02362, over 1441234.92 frames. ], batch size: 93, lr: 4.02e-03, grad_scale: 8.0 +2023-03-21 10:17:14,688 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 10:17:22,880 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1140, 4.6458, 4.7766, 4.6648, 4.6599, 4.2044, 4.7516, 4.5671], + device='cuda:1'), covar=tensor([0.0534, 0.0377, 0.0306, 0.0455, 0.0311, 0.0460, 0.0307, 0.0462], + device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0263, 0.0201, 0.0204, 0.0160, 0.0231, 0.0214, 0.0149], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:17:24,294 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 10:17:24,670 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 +2023-03-21 10:17:24,861 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=114831.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:17:35,697 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9988, 2.2514, 2.2836, 2.2008, 2.1970, 2.3025, 2.1066, 1.8479], + device='cuda:1'), covar=tensor([0.0776, 0.0673, 0.0449, 0.0307, 0.0503, 0.0399, 0.0342, 0.0265], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0039, 0.0041, 0.0039, 0.0037, 0.0037, 0.0042, 0.0042], + device='cuda:1'), out_proj_covar=tensor([1.0369e-04, 1.0181e-04, 1.0177e-04, 9.8873e-05, 9.8211e-05, 9.6444e-05, + 1.0553e-04, 1.0636e-04], device='cuda:1') +2023-03-21 10:17:36,016 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.665e+02 1.955e+02 2.157e+02 5.091e+02, threshold=3.909e+02, percent-clipped=1.0 +2023-03-21 10:17:40,524 INFO [train.py:901] (1/2) Epoch 41, batch 1900, loss[loss=0.1372, simple_loss=0.2233, pruned_loss=0.02555, over 7279.00 frames. ], tot_loss[loss=0.1289, simple_loss=0.2105, pruned_loss=0.02366, over 1440784.00 frames. ], batch size: 77, lr: 4.02e-03, grad_scale: 8.0 +2023-03-21 10:17:41,066 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 10:17:46,363 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114871.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:18:06,217 INFO [train.py:901] (1/2) Epoch 41, batch 1950, loss[loss=0.1307, simple_loss=0.2057, pruned_loss=0.0279, over 7304.00 frames. ], tot_loss[loss=0.129, simple_loss=0.2104, pruned_loss=0.02385, over 1441263.80 frames. ], batch size: 49, lr: 4.02e-03, grad_scale: 8.0 +2023-03-21 10:18:06,254 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 10:18:16,434 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6835, 2.3872, 2.4238, 3.8981, 2.0822, 3.6148, 1.4875, 3.5336], + device='cuda:1'), covar=tensor([0.0201, 0.1548, 0.1913, 0.0248, 0.3621, 0.0297, 0.1257, 0.0361], + device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0251, 0.0262, 0.0213, 0.0253, 0.0221, 0.0226, 0.0232], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 10:18:16,816 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 10:18:16,940 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114932.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:18:21,750 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 10:18:22,246 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 10:18:27,917 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.257e+02 1.719e+02 2.045e+02 2.389e+02 4.701e+02, threshold=4.090e+02, percent-clipped=2.0 +2023-03-21 10:18:31,232 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2905, 2.5290, 2.5924, 2.3280, 2.3196, 2.3273, 2.3099, 2.0123], + device='cuda:1'), covar=tensor([0.0602, 0.0523, 0.0331, 0.0322, 0.0743, 0.0684, 0.0267, 0.0272], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0039, 0.0041, 0.0039, 0.0037, 0.0037, 0.0042, 0.0042], + device='cuda:1'), out_proj_covar=tensor([1.0369e-04, 1.0181e-04, 1.0207e-04, 9.9410e-05, 9.8271e-05, 9.6909e-05, + 1.0580e-04, 1.0656e-04], device='cuda:1') +2023-03-21 10:18:32,589 INFO [train.py:901] (1/2) Epoch 41, batch 2000, loss[loss=0.1191, simple_loss=0.2049, pruned_loss=0.01661, over 7237.00 frames. ], tot_loss[loss=0.1295, simple_loss=0.211, pruned_loss=0.024, over 1442103.67 frames. ], batch size: 89, lr: 4.01e-03, grad_scale: 8.0 +2023-03-21 10:18:39,249 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 10:18:40,546 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 +2023-03-21 10:18:49,404 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 10:18:54,987 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2035, 2.6015, 2.7683, 2.6089, 2.3813, 2.5636, 2.3273, 2.0793], + device='cuda:1'), covar=tensor([0.0584, 0.0436, 0.0331, 0.0232, 0.0727, 0.0418, 0.0338, 0.0324], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0039, 0.0041, 0.0039, 0.0037, 0.0037, 0.0042, 0.0042], + device='cuda:1'), out_proj_covar=tensor([1.0339e-04, 1.0160e-04, 1.0175e-04, 9.8964e-05, 9.7915e-05, 9.6645e-05, + 1.0568e-04, 1.0619e-04], device='cuda:1') +2023-03-21 10:18:56,013 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4672, 1.7420, 1.5720, 1.6529, 1.8585, 1.7361, 1.5996, 1.4064], + device='cuda:1'), covar=tensor([0.0177, 0.0183, 0.0207, 0.0191, 0.0115, 0.0109, 0.0126, 0.0209], + device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0037, 0.0035, 0.0037, 0.0035, 0.0033, 0.0037, 0.0045], + device='cuda:1'), out_proj_covar=tensor([4.2202e-05, 4.1065e-05, 3.9564e-05, 4.0823e-05, 3.8983e-05, 3.6919e-05, + 4.1593e-05, 4.9601e-05], device='cuda:1') +2023-03-21 10:18:57,875 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 10:18:58,358 INFO [train.py:901] (1/2) Epoch 41, batch 2050, loss[loss=0.1482, simple_loss=0.2319, pruned_loss=0.03231, over 7340.00 frames. ], tot_loss[loss=0.1297, simple_loss=0.2112, pruned_loss=0.0241, over 1443411.52 frames. ], batch size: 54, lr: 4.01e-03, grad_scale: 8.0 +2023-03-21 10:19:20,704 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 1.710e+02 1.996e+02 2.355e+02 4.786e+02, threshold=3.993e+02, percent-clipped=2.0 +2023-03-21 10:19:24,728 INFO [train.py:901] (1/2) Epoch 41, batch 2100, loss[loss=0.1239, simple_loss=0.2121, pruned_loss=0.0179, over 7302.00 frames. ], tot_loss[loss=0.1299, simple_loss=0.2116, pruned_loss=0.02407, over 1443032.58 frames. ], batch size: 80, lr: 4.01e-03, grad_scale: 8.0 +2023-03-21 10:19:30,519 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115072.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:19:31,992 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 10:19:34,604 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 10:19:37,730 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115086.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:19:46,180 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6706, 3.1284, 2.7472, 2.9494, 3.0432, 2.8445, 3.0439, 2.8837], + device='cuda:1'), covar=tensor([0.0852, 0.0589, 0.1067, 0.1669, 0.0955, 0.0648, 0.0680, 0.1090], + device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0059, 0.0068, 0.0060, 0.0057, 0.0063, 0.0057, 0.0054], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 10:19:50,075 INFO [train.py:901] (1/2) Epoch 41, batch 2150, loss[loss=0.1405, simple_loss=0.2197, pruned_loss=0.03061, over 7247.00 frames. ], tot_loss[loss=0.1294, simple_loss=0.2113, pruned_loss=0.02371, over 1442020.48 frames. ], batch size: 47, lr: 4.01e-03, grad_scale: 8.0 +2023-03-21 10:20:09,490 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115147.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:20:12,343 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.680e+02 1.975e+02 2.353e+02 5.510e+02, threshold=3.949e+02, percent-clipped=1.0 +2023-03-21 10:20:16,348 INFO [train.py:901] (1/2) Epoch 41, batch 2200, loss[loss=0.1362, simple_loss=0.2135, pruned_loss=0.02939, over 7271.00 frames. ], tot_loss[loss=0.1293, simple_loss=0.2113, pruned_loss=0.02362, over 1444011.08 frames. ], batch size: 70, lr: 4.01e-03, grad_scale: 8.0 +2023-03-21 10:20:17,952 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 10:20:21,587 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0938, 3.4360, 4.1911, 4.0817, 4.1568, 4.2424, 4.2987, 4.1496], + device='cuda:1'), covar=tensor([0.0032, 0.0113, 0.0028, 0.0028, 0.0032, 0.0027, 0.0027, 0.0041], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0070, 0.0058, 0.0057, 0.0055, 0.0060, 0.0049, 0.0077], + device='cuda:1'), out_proj_covar=tensor([8.1701e-05, 1.4171e-04, 1.0354e-04, 9.7122e-05, 9.3572e-05, 1.0401e-04, + 9.2604e-05, 1.4072e-04], device='cuda:1') +2023-03-21 10:20:38,267 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.08 vs. limit=5.0 +2023-03-21 10:20:42,139 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5671, 4.1294, 3.8961, 4.5447, 4.2753, 4.4312, 3.9569, 4.0181], + device='cuda:1'), covar=tensor([0.0977, 0.2618, 0.2458, 0.1157, 0.1066, 0.1250, 0.0968, 0.1286], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0406, 0.0305, 0.0321, 0.0237, 0.0380, 0.0236, 0.0284], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 10:20:43,075 INFO [train.py:901] (1/2) Epoch 41, batch 2250, loss[loss=0.101, simple_loss=0.1651, pruned_loss=0.01845, over 5808.00 frames. ], tot_loss[loss=0.1286, simple_loss=0.2106, pruned_loss=0.02328, over 1441910.83 frames. ], batch size: 25, lr: 4.01e-03, grad_scale: 8.0 +2023-03-21 10:20:47,240 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8885, 3.8419, 3.0677, 3.5249, 2.6994, 2.1999, 1.7572, 3.8837], + device='cuda:1'), covar=tensor([0.0052, 0.0054, 0.0174, 0.0082, 0.0218, 0.0585, 0.0704, 0.0062], + device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0094, 0.0113, 0.0095, 0.0131, 0.0135, 0.0128, 0.0106], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 10:20:51,200 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115227.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:20:52,172 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 10:20:52,186 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 10:21:04,394 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.338e+02 1.786e+02 2.050e+02 2.347e+02 4.305e+02, threshold=4.099e+02, percent-clipped=1.0 +2023-03-21 10:21:04,938 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 10:21:08,410 INFO [train.py:901] (1/2) Epoch 41, batch 2300, loss[loss=0.1402, simple_loss=0.2293, pruned_loss=0.02558, over 6688.00 frames. ], tot_loss[loss=0.1288, simple_loss=0.2107, pruned_loss=0.02346, over 1440967.28 frames. ], batch size: 107, lr: 4.01e-03, grad_scale: 16.0 +2023-03-21 10:21:34,777 INFO [train.py:901] (1/2) Epoch 41, batch 2350, loss[loss=0.1287, simple_loss=0.2109, pruned_loss=0.0232, over 7340.00 frames. ], tot_loss[loss=0.1289, simple_loss=0.2108, pruned_loss=0.02351, over 1443355.62 frames. ], batch size: 61, lr: 4.01e-03, grad_scale: 16.0 +2023-03-21 10:21:51,992 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 10:21:56,116 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.249e+02 1.722e+02 2.004e+02 2.418e+02 5.558e+02, threshold=4.008e+02, percent-clipped=3.0 +2023-03-21 10:21:58,711 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 10:22:00,257 INFO [train.py:901] (1/2) Epoch 41, batch 2400, loss[loss=0.1242, simple_loss=0.2087, pruned_loss=0.01986, over 7238.00 frames. ], tot_loss[loss=0.129, simple_loss=0.2108, pruned_loss=0.02364, over 1443242.01 frames. ], batch size: 55, lr: 4.01e-03, grad_scale: 16.0 +2023-03-21 10:22:05,670 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115370.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:22:06,145 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115371.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:22:06,675 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115372.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:22:09,777 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 10:22:12,844 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 10:22:13,930 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9641, 2.3025, 2.3114, 2.2211, 2.0673, 2.0175, 1.9223, 1.6830], + device='cuda:1'), covar=tensor([0.0647, 0.0487, 0.0423, 0.0299, 0.0528, 0.0724, 0.0476, 0.0569], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0039, 0.0040, 0.0038, 0.0037, 0.0037, 0.0042, 0.0042], + device='cuda:1'), out_proj_covar=tensor([1.0184e-04, 1.0050e-04, 1.0084e-04, 9.7865e-05, 9.6769e-05, 9.5952e-05, + 1.0464e-04, 1.0507e-04], device='cuda:1') +2023-03-21 10:22:27,157 INFO [train.py:901] (1/2) Epoch 41, batch 2450, loss[loss=0.1366, simple_loss=0.2168, pruned_loss=0.02822, over 7334.00 frames. ], tot_loss[loss=0.1294, simple_loss=0.2111, pruned_loss=0.02385, over 1444670.15 frames. ], batch size: 54, lr: 4.01e-03, grad_scale: 16.0 +2023-03-21 10:22:31,873 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=115420.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:22:37,372 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115431.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:22:37,873 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115432.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:22:38,244 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 10:22:42,791 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115442.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:22:43,307 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2794, 3.3958, 4.3122, 4.2764, 4.2644, 4.3311, 4.2998, 4.2327], + device='cuda:1'), covar=tensor([0.0028, 0.0119, 0.0031, 0.0024, 0.0027, 0.0031, 0.0027, 0.0048], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0070, 0.0058, 0.0057, 0.0055, 0.0060, 0.0049, 0.0077], + device='cuda:1'), out_proj_covar=tensor([8.1880e-05, 1.4160e-04, 1.0356e-04, 9.6488e-05, 9.2953e-05, 1.0399e-04, + 9.2342e-05, 1.4025e-04], device='cuda:1') +2023-03-21 10:22:48,839 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.303e+02 1.715e+02 1.917e+02 2.250e+02 3.232e+02, threshold=3.834e+02, percent-clipped=0.0 +2023-03-21 10:22:53,620 INFO [train.py:901] (1/2) Epoch 41, batch 2500, loss[loss=0.1192, simple_loss=0.2084, pruned_loss=0.01501, over 7255.00 frames. ], tot_loss[loss=0.1295, simple_loss=0.2112, pruned_loss=0.02387, over 1444878.82 frames. ], batch size: 89, lr: 4.01e-03, grad_scale: 16.0 +2023-03-21 10:23:03,581 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 10:23:18,444 INFO [train.py:901] (1/2) Epoch 41, batch 2550, loss[loss=0.1354, simple_loss=0.2219, pruned_loss=0.02449, over 7249.00 frames. ], tot_loss[loss=0.1301, simple_loss=0.2116, pruned_loss=0.02424, over 1444829.94 frames. ], batch size: 89, lr: 4.00e-03, grad_scale: 8.0 +2023-03-21 10:23:26,740 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115527.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:23:41,547 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.217e+02 1.617e+02 1.970e+02 2.385e+02 4.458e+02, threshold=3.941e+02, percent-clipped=1.0 +2023-03-21 10:23:45,023 INFO [train.py:901] (1/2) Epoch 41, batch 2600, loss[loss=0.1288, simple_loss=0.212, pruned_loss=0.02277, over 7267.00 frames. ], tot_loss[loss=0.1298, simple_loss=0.2113, pruned_loss=0.02413, over 1442289.50 frames. ], batch size: 57, lr: 4.00e-03, grad_scale: 8.0 +2023-03-21 10:23:46,623 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0369, 3.4970, 4.0307, 3.9973, 4.0490, 4.0570, 4.1741, 3.8873], + device='cuda:1'), covar=tensor([0.0031, 0.0096, 0.0028, 0.0031, 0.0028, 0.0032, 0.0027, 0.0045], + device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0070, 0.0058, 0.0056, 0.0055, 0.0060, 0.0049, 0.0076], + device='cuda:1'), out_proj_covar=tensor([8.1225e-05, 1.4042e-04, 1.0297e-04, 9.5981e-05, 9.2125e-05, 1.0349e-04, + 9.1655e-05, 1.3903e-04], device='cuda:1') +2023-03-21 10:23:52,035 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=115575.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:24:02,660 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 +2023-03-21 10:24:03,906 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3256, 2.3606, 2.3287, 3.4881, 1.9080, 3.3322, 1.4168, 3.2203], + device='cuda:1'), covar=tensor([0.0260, 0.1516, 0.2026, 0.0282, 0.4056, 0.0333, 0.1345, 0.0475], + device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0247, 0.0259, 0.0213, 0.0251, 0.0218, 0.0225, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 10:24:09,934 INFO [train.py:901] (1/2) Epoch 41, batch 2650, loss[loss=0.1264, simple_loss=0.212, pruned_loss=0.02036, over 7356.00 frames. ], tot_loss[loss=0.1298, simple_loss=0.2115, pruned_loss=0.02411, over 1441245.76 frames. ], batch size: 51, lr: 4.00e-03, grad_scale: 8.0 +2023-03-21 10:24:22,739 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-21 10:24:23,065 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7097, 1.6253, 2.1501, 2.3233, 2.0831, 2.1278, 1.9974, 2.4163], + device='cuda:1'), covar=tensor([0.2165, 0.4826, 0.1518, 0.1101, 0.1950, 0.3274, 0.2285, 0.2061], + device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0082, 0.0074, 0.0067, 0.0068, 0.0067, 0.0108, 0.0071], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:24:27,077 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8612, 3.2889, 2.9589, 3.2190, 3.1569, 2.8684, 3.1487, 2.9260], + device='cuda:1'), covar=tensor([0.0896, 0.0601, 0.0728, 0.1218, 0.0704, 0.0541, 0.1036, 0.0946], + device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0060, 0.0068, 0.0061, 0.0057, 0.0063, 0.0057, 0.0055], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:24:31,803 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.159e+02 1.689e+02 1.943e+02 2.398e+02 3.998e+02, threshold=3.887e+02, percent-clipped=1.0 +2023-03-21 10:24:34,766 INFO [train.py:901] (1/2) Epoch 41, batch 2700, loss[loss=0.1299, simple_loss=0.2208, pruned_loss=0.01948, over 7253.00 frames. ], tot_loss[loss=0.1295, simple_loss=0.2108, pruned_loss=0.02413, over 1441072.29 frames. ], batch size: 57, lr: 4.00e-03, grad_scale: 4.0 +2023-03-21 10:24:47,950 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 +2023-03-21 10:24:59,557 INFO [train.py:901] (1/2) Epoch 41, batch 2750, loss[loss=0.1108, simple_loss=0.2007, pruned_loss=0.01047, over 7302.00 frames. ], tot_loss[loss=0.1289, simple_loss=0.2104, pruned_loss=0.02376, over 1440476.71 frames. ], batch size: 86, lr: 4.00e-03, grad_scale: 4.0 +2023-03-21 10:25:00,149 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1022, 3.7549, 3.7756, 3.7487, 3.7414, 3.6752, 3.9413, 3.5659], + device='cuda:1'), covar=tensor([0.0125, 0.0201, 0.0129, 0.0187, 0.0453, 0.0123, 0.0169, 0.0187], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0107, 0.0108, 0.0093, 0.0184, 0.0112, 0.0111, 0.0117], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 10:25:00,367 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 +2023-03-21 10:25:07,116 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115726.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:25:07,587 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115727.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:25:11,937 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.1568, 4.6854, 4.5049, 5.1187, 4.8971, 5.0488, 4.5759, 4.6930], + device='cuda:1'), covar=tensor([0.0845, 0.2534, 0.2313, 0.1020, 0.0986, 0.1163, 0.0914, 0.1197], + device='cuda:1'), in_proj_covar=tensor([0.0267, 0.0402, 0.0305, 0.0318, 0.0236, 0.0379, 0.0237, 0.0282], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 10:25:14,883 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115742.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:25:21,067 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.281e+02 1.781e+02 2.069e+02 2.360e+02 3.753e+02, threshold=4.137e+02, percent-clipped=0.0 +2023-03-21 10:25:23,957 INFO [train.py:901] (1/2) Epoch 41, batch 2800, loss[loss=0.1318, simple_loss=0.2143, pruned_loss=0.02461, over 7266.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.211, pruned_loss=0.02412, over 1442223.09 frames. ], batch size: 64, lr: 4.00e-03, grad_scale: 8.0 +2023-03-21 10:25:48,104 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 10:25:48,157 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 10:25:48,181 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 10:25:48,480 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 10:25:48,647 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 10:25:48,662 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 10:25:48,673 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 10:25:48,775 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 10:25:48,849 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 10:25:48,864 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 10:25:48,880 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 10:25:49,039 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 10:25:49,046 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 10:25:49,059 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 10:25:49,540 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 10:25:49,543 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 10:25:49,600 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 10:25:56,686 INFO [train.py:901] (1/2) Epoch 42, batch 0, loss[loss=0.1257, simple_loss=0.2119, pruned_loss=0.01971, over 7265.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.2119, pruned_loss=0.01971, over 7265.00 frames. ], batch size: 70, lr: 3.95e-03, grad_scale: 8.0 +2023-03-21 10:25:56,686 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 10:26:06,813 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9058, 3.2255, 2.4187, 3.5458, 2.9729, 3.0494, 1.5862, 2.5644], + device='cuda:1'), covar=tensor([0.0457, 0.0808, 0.2885, 0.0524, 0.0583, 0.0597, 0.4289, 0.1986], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0258, 0.0277, 0.0267, 0.0270, 0.0264, 0.0232, 0.0254], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 10:26:12,712 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.4150, 5.1067, 4.7538, 5.4929, 5.1538, 5.4710, 5.1657, 5.1593], + device='cuda:1'), covar=tensor([0.0539, 0.1928, 0.1733, 0.0788, 0.0761, 0.0681, 0.0443, 0.0787], + device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0405, 0.0307, 0.0321, 0.0238, 0.0381, 0.0238, 0.0284], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 10:26:19,861 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6047, 1.9898, 1.5198, 1.9287, 2.0094, 1.8497, 1.7794, 1.5168], + device='cuda:1'), covar=tensor([0.0184, 0.0153, 0.0277, 0.0167, 0.0105, 0.0158, 0.0223, 0.0232], + device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0038, 0.0036, 0.0038, 0.0036, 0.0034, 0.0038, 0.0046], + device='cuda:1'), out_proj_covar=tensor([4.3531e-05, 4.2313e-05, 4.1211e-05, 4.1674e-05, 4.0039e-05, 3.8308e-05, + 4.2724e-05, 5.0491e-05], device='cuda:1') +2023-03-21 10:26:22,420 INFO [train.py:935] (1/2) Epoch 42, validation: loss=0.166, simple_loss=0.2576, pruned_loss=0.03718, over 1622729.00 frames. +2023-03-21 10:26:22,421 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 10:26:25,060 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=115790.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:26:29,971 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 10:26:41,189 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 10:26:47,220 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 10:26:49,220 INFO [train.py:901] (1/2) Epoch 42, batch 50, loss[loss=0.1361, simple_loss=0.2231, pruned_loss=0.02453, over 7284.00 frames. ], tot_loss[loss=0.13, simple_loss=0.2119, pruned_loss=0.02402, over 324472.28 frames. ], batch size: 57, lr: 3.95e-03, grad_scale: 8.0 +2023-03-21 10:26:49,231 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 10:26:52,196 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 10:26:59,262 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.182e+02 1.683e+02 1.944e+02 2.243e+02 4.177e+02, threshold=3.887e+02, percent-clipped=1.0 +2023-03-21 10:27:09,387 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 10:27:10,444 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 10:27:13,056 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9044, 1.7032, 2.3084, 2.4003, 2.1909, 2.2959, 2.2124, 2.3479], + device='cuda:1'), covar=tensor([0.4633, 0.3934, 0.3944, 0.1657, 0.3140, 0.2680, 0.2713, 0.4267], + device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0083, 0.0075, 0.0068, 0.0068, 0.0068, 0.0110, 0.0072], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:27:14,893 INFO [train.py:901] (1/2) Epoch 42, batch 100, loss[loss=0.1317, simple_loss=0.2114, pruned_loss=0.026, over 7261.00 frames. ], tot_loss[loss=0.1283, simple_loss=0.2095, pruned_loss=0.02351, over 572618.94 frames. ], batch size: 57, lr: 3.95e-03, grad_scale: 8.0 +2023-03-21 10:27:15,533 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6802, 1.5105, 1.9305, 2.1338, 1.7568, 1.9431, 1.6633, 2.1590], + device='cuda:1'), covar=tensor([0.3542, 0.3859, 0.2085, 0.1286, 0.1430, 0.2738, 0.2393, 0.2912], + device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0083, 0.0075, 0.0068, 0.0068, 0.0068, 0.0110, 0.0072], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:27:19,170 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115892.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 10:27:40,728 INFO [train.py:901] (1/2) Epoch 42, batch 150, loss[loss=0.1252, simple_loss=0.2022, pruned_loss=0.02411, over 7251.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2101, pruned_loss=0.02337, over 767161.75 frames. ], batch size: 64, lr: 3.95e-03, grad_scale: 8.0 +2023-03-21 10:27:49,874 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115953.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 10:27:50,693 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.221e+02 1.790e+02 2.023e+02 2.356e+02 3.555e+02, threshold=4.045e+02, percent-clipped=0.0 +2023-03-21 10:28:07,101 INFO [train.py:901] (1/2) Epoch 42, batch 200, loss[loss=0.1073, simple_loss=0.1896, pruned_loss=0.01248, over 7144.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2098, pruned_loss=0.02316, over 915086.70 frames. ], batch size: 41, lr: 3.95e-03, grad_scale: 8.0 +2023-03-21 10:28:12,638 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 10:28:21,030 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 10:28:25,674 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116014.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:28:27,059 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 10:28:31,672 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116026.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:28:32,143 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116027.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:28:35,361 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-03-21 10:28:35,998 INFO [train.py:901] (1/2) Epoch 42, batch 250, loss[loss=0.1236, simple_loss=0.197, pruned_loss=0.02514, over 7200.00 frames. ], tot_loss[loss=0.1285, simple_loss=0.2102, pruned_loss=0.02346, over 1033982.76 frames. ], batch size: 39, lr: 3.95e-03, grad_scale: 8.0 +2023-03-21 10:28:40,489 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 10:28:44,807 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-03-21 10:28:46,482 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.767e+02 2.086e+02 2.345e+02 3.684e+02, threshold=4.171e+02, percent-clipped=0.0 +2023-03-21 10:28:47,261 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-03-21 10:28:56,705 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=116074.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:28:57,204 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=116075.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:28:57,305 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116075.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 10:29:00,574 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 10:29:02,016 INFO [train.py:901] (1/2) Epoch 42, batch 300, loss[loss=0.1594, simple_loss=0.2348, pruned_loss=0.04203, over 7343.00 frames. ], tot_loss[loss=0.1287, simple_loss=0.21, pruned_loss=0.02376, over 1125110.60 frames. ], batch size: 54, lr: 3.95e-03, grad_scale: 8.0 +2023-03-21 10:29:08,742 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 10:29:11,934 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116105.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:29:17,445 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116116.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:29:26,750 INFO [train.py:901] (1/2) Epoch 42, batch 350, loss[loss=0.1262, simple_loss=0.2041, pruned_loss=0.02417, over 7230.00 frames. ], tot_loss[loss=0.1293, simple_loss=0.2106, pruned_loss=0.02404, over 1196845.59 frames. ], batch size: 45, lr: 3.95e-03, grad_scale: 8.0 +2023-03-21 10:29:37,068 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.345e+02 1.856e+02 2.238e+02 2.554e+02 4.573e+02, threshold=4.477e+02, percent-clipped=1.0 +2023-03-21 10:29:43,297 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 10:29:43,459 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116166.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:29:48,963 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116177.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:29:52,560 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3773, 2.8601, 2.3033, 3.3069, 3.1106, 2.9543, 2.6673, 2.9117], + device='cuda:1'), covar=tensor([0.2137, 0.1017, 0.3727, 0.0661, 0.0343, 0.0288, 0.0383, 0.0528], + device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0230, 0.0244, 0.0255, 0.0199, 0.0202, 0.0219, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 10:29:52,896 INFO [train.py:901] (1/2) Epoch 42, batch 400, loss[loss=0.1128, simple_loss=0.1991, pruned_loss=0.01328, over 7340.00 frames. ], tot_loss[loss=0.1293, simple_loss=0.2104, pruned_loss=0.02409, over 1249733.67 frames. ], batch size: 44, lr: 3.95e-03, grad_scale: 8.0 +2023-03-21 10:30:10,577 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 +2023-03-21 10:30:18,830 INFO [train.py:901] (1/2) Epoch 42, batch 450, loss[loss=0.1335, simple_loss=0.2182, pruned_loss=0.02433, over 7315.00 frames. ], tot_loss[loss=0.1291, simple_loss=0.2102, pruned_loss=0.02399, over 1293480.93 frames. ], batch size: 83, lr: 3.94e-03, grad_scale: 8.0 +2023-03-21 10:30:19,528 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7110, 2.6486, 2.5758, 3.7098, 2.0115, 3.5485, 1.4745, 3.4042], + device='cuda:1'), covar=tensor([0.0244, 0.1272, 0.1859, 0.0271, 0.4010, 0.0317, 0.1324, 0.0348], + device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0246, 0.0259, 0.0214, 0.0251, 0.0219, 0.0226, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 10:30:20,579 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0242, 2.8182, 3.0915, 3.0284, 2.7384, 2.8046, 3.3153, 2.3302], + device='cuda:1'), covar=tensor([0.0604, 0.0661, 0.0891, 0.0754, 0.0798, 0.1158, 0.0970, 0.2783], + device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0335, 0.0271, 0.0352, 0.0284, 0.0285, 0.0348, 0.0242], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:30:25,515 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 10:30:25,534 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 10:30:26,096 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116248.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 10:30:29,792 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 +2023-03-21 10:30:29,976 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.379e+02 1.720e+02 2.089e+02 2.447e+02 6.240e+02, threshold=4.178e+02, percent-clipped=2.0 +2023-03-21 10:30:44,580 INFO [train.py:901] (1/2) Epoch 42, batch 500, loss[loss=0.1417, simple_loss=0.2214, pruned_loss=0.03097, over 6548.00 frames. ], tot_loss[loss=0.1287, simple_loss=0.2097, pruned_loss=0.02382, over 1325155.42 frames. ], batch size: 106, lr: 3.94e-03, grad_scale: 4.0 +2023-03-21 10:30:57,706 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 10:30:59,090 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 10:31:00,101 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 10:31:00,703 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7154, 5.2241, 5.2661, 5.2596, 4.9996, 4.7695, 5.3060, 5.1148], + device='cuda:1'), covar=tensor([0.0390, 0.0329, 0.0347, 0.0407, 0.0300, 0.0355, 0.0310, 0.0384], + device='cuda:1'), in_proj_covar=tensor([0.0220, 0.0261, 0.0203, 0.0204, 0.0158, 0.0231, 0.0214, 0.0149], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:31:01,654 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 10:31:04,266 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116323.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:31:07,762 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 10:31:10,709 INFO [train.py:901] (1/2) Epoch 42, batch 550, loss[loss=0.1292, simple_loss=0.214, pruned_loss=0.02214, over 7289.00 frames. ], tot_loss[loss=0.129, simple_loss=0.2102, pruned_loss=0.02391, over 1353447.28 frames. ], batch size: 66, lr: 3.94e-03, grad_scale: 4.0 +2023-03-21 10:31:19,561 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 10:31:21,545 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.746e+02 2.072e+02 2.428e+02 3.204e+02, threshold=4.144e+02, percent-clipped=0.0 +2023-03-21 10:31:27,603 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 10:31:28,737 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116370.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 10:31:31,122 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 10:31:35,759 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116384.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:31:36,111 INFO [train.py:901] (1/2) Epoch 42, batch 600, loss[loss=0.09882, simple_loss=0.1767, pruned_loss=0.01048, over 7018.00 frames. ], tot_loss[loss=0.1289, simple_loss=0.21, pruned_loss=0.02394, over 1371797.37 frames. ], batch size: 35, lr: 3.94e-03, grad_scale: 4.0 +2023-03-21 10:31:37,626 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 10:31:54,400 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4915, 3.6971, 2.7377, 3.9827, 3.2020, 3.6593, 1.9436, 2.7978], + device='cuda:1'), covar=tensor([0.0480, 0.0890, 0.2271, 0.0591, 0.0507, 0.0699, 0.3583, 0.1746], + device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0260, 0.0278, 0.0269, 0.0272, 0.0265, 0.0233, 0.0257], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 10:31:54,736 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 10:32:02,795 INFO [train.py:901] (1/2) Epoch 42, batch 650, loss[loss=0.1491, simple_loss=0.2301, pruned_loss=0.03405, over 7348.00 frames. ], tot_loss[loss=0.1291, simple_loss=0.2104, pruned_loss=0.02391, over 1388748.25 frames. ], batch size: 51, lr: 3.94e-03, grad_scale: 4.0 +2023-03-21 10:32:03,761 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 10:32:08,269 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116446.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:32:13,281 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.271e+02 1.770e+02 2.062e+02 2.474e+02 7.007e+02, threshold=4.123e+02, percent-clipped=2.0 +2023-03-21 10:32:15,953 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116461.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:32:20,983 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 10:32:21,542 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116472.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:32:27,958 INFO [train.py:901] (1/2) Epoch 42, batch 700, loss[loss=0.1384, simple_loss=0.2213, pruned_loss=0.02778, over 7294.00 frames. ], tot_loss[loss=0.1291, simple_loss=0.21, pruned_loss=0.02409, over 1400936.71 frames. ], batch size: 57, lr: 3.94e-03, grad_scale: 4.0 +2023-03-21 10:32:29,092 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 10:32:30,795 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 +2023-03-21 10:32:39,413 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116505.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:32:40,453 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116507.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 10:32:54,356 INFO [train.py:901] (1/2) Epoch 42, batch 750, loss[loss=0.1143, simple_loss=0.2019, pruned_loss=0.01331, over 7347.00 frames. ], tot_loss[loss=0.1283, simple_loss=0.2095, pruned_loss=0.02356, over 1411259.50 frames. ], batch size: 73, lr: 3.94e-03, grad_scale: 4.0 +2023-03-21 10:32:54,362 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 10:32:54,876 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 10:33:00,943 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116548.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 10:33:04,818 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 1.798e+02 2.124e+02 2.350e+02 4.773e+02, threshold=4.247e+02, percent-clipped=1.0 +2023-03-21 10:33:08,824 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 10:33:09,909 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116566.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:33:12,867 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 10:33:19,503 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 10:33:19,999 INFO [train.py:901] (1/2) Epoch 42, batch 800, loss[loss=0.1198, simple_loss=0.2076, pruned_loss=0.01603, over 7312.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2091, pruned_loss=0.02359, over 1415529.02 frames. ], batch size: 80, lr: 3.94e-03, grad_scale: 8.0 +2023-03-21 10:33:20,513 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 10:33:26,290 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=116596.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 10:33:31,703 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 10:33:36,493 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-03-21 10:33:37,733 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116618.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:33:44,227 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7507, 1.6893, 2.0938, 2.1001, 1.9544, 2.0758, 1.8042, 2.2795], + device='cuda:1'), covar=tensor([0.2353, 0.2749, 0.1011, 0.0875, 0.2277, 0.2203, 0.2157, 0.2039], + device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0082, 0.0075, 0.0068, 0.0068, 0.0067, 0.0109, 0.0072], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:33:46,115 INFO [train.py:901] (1/2) Epoch 42, batch 850, loss[loss=0.102, simple_loss=0.1714, pruned_loss=0.01628, over 6970.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2096, pruned_loss=0.0236, over 1422380.28 frames. ], batch size: 35, lr: 3.94e-03, grad_scale: 8.0 +2023-03-21 10:33:50,677 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 10:33:50,689 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 10:33:56,154 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 10:33:56,614 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.288e+02 1.744e+02 2.049e+02 2.430e+02 4.238e+02, threshold=4.097e+02, percent-clipped=0.0 +2023-03-21 10:34:00,120 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 10:34:04,265 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116670.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 10:34:09,208 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:34:09,261 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:34:12,097 INFO [train.py:901] (1/2) Epoch 42, batch 900, loss[loss=0.1046, simple_loss=0.1865, pruned_loss=0.01135, over 7345.00 frames. ], tot_loss[loss=0.1287, simple_loss=0.2096, pruned_loss=0.02385, over 1426177.20 frames. ], batch size: 44, lr: 3.94e-03, grad_scale: 8.0 +2023-03-21 10:34:21,786 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116704.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:34:28,577 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=116718.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:34:37,081 INFO [train.py:901] (1/2) Epoch 42, batch 950, loss[loss=0.1402, simple_loss=0.2208, pruned_loss=0.02976, over 7275.00 frames. ], tot_loss[loss=0.1285, simple_loss=0.2095, pruned_loss=0.02377, over 1431546.53 frames. ], batch size: 70, lr: 3.94e-03, grad_scale: 8.0 +2023-03-21 10:34:37,082 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 10:34:48,295 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.041e+02 1.811e+02 2.090e+02 2.390e+02 5.316e+02, threshold=4.179e+02, percent-clipped=3.0 +2023-03-21 10:34:51,631 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116761.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:34:53,719 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116765.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:34:57,237 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116772.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:35:02,126 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 10:35:03,563 INFO [train.py:901] (1/2) Epoch 42, batch 1000, loss[loss=0.1294, simple_loss=0.2085, pruned_loss=0.02517, over 7272.00 frames. ], tot_loss[loss=0.1285, simple_loss=0.2097, pruned_loss=0.02369, over 1434267.22 frames. ], batch size: 47, lr: 3.94e-03, grad_scale: 8.0 +2023-03-21 10:35:12,529 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116802.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 10:35:12,569 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116802.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:35:16,084 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=116809.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:35:21,583 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 10:35:21,639 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=116820.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:35:29,790 INFO [train.py:901] (1/2) Epoch 42, batch 1050, loss[loss=0.1064, simple_loss=0.1673, pruned_loss=0.02274, over 5945.00 frames. ], tot_loss[loss=0.1289, simple_loss=0.2101, pruned_loss=0.02381, over 1435781.73 frames. ], batch size: 25, lr: 3.93e-03, grad_scale: 8.0 +2023-03-21 10:35:30,914 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4971, 4.0714, 4.1039, 4.1110, 4.0695, 3.9958, 4.2814, 3.5999], + device='cuda:1'), covar=tensor([0.0138, 0.0158, 0.0123, 0.0174, 0.0463, 0.0136, 0.0156, 0.0262], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0105, 0.0105, 0.0090, 0.0180, 0.0109, 0.0108, 0.0116], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 10:35:40,888 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.209e+02 1.773e+02 2.097e+02 2.514e+02 5.231e+02, threshold=4.194e+02, percent-clipped=1.0 +2023-03-21 10:35:43,597 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116861.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:35:44,058 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 10:35:44,709 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116863.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:35:48,138 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 10:35:55,649 INFO [train.py:901] (1/2) Epoch 42, batch 1100, loss[loss=0.1413, simple_loss=0.2211, pruned_loss=0.03075, over 7309.00 frames. ], tot_loss[loss=0.1286, simple_loss=0.2098, pruned_loss=0.02371, over 1436606.64 frames. ], batch size: 49, lr: 3.93e-03, grad_scale: 8.0 +2023-03-21 10:35:56,272 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2537, 2.2525, 2.2745, 3.3965, 1.8513, 3.3768, 1.3145, 3.1205], + device='cuda:1'), covar=tensor([0.0306, 0.1524, 0.2058, 0.0345, 0.4490, 0.0380, 0.1401, 0.0491], + device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0243, 0.0256, 0.0211, 0.0249, 0.0217, 0.0223, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 10:36:16,908 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 10:36:16,923 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 10:36:21,979 INFO [train.py:901] (1/2) Epoch 42, batch 1150, loss[loss=0.1338, simple_loss=0.2117, pruned_loss=0.02791, over 7331.00 frames. ], tot_loss[loss=0.1287, simple_loss=0.2102, pruned_loss=0.02357, over 1439279.05 frames. ], batch size: 54, lr: 3.93e-03, grad_scale: 8.0 +2023-03-21 10:36:29,641 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 10:36:30,140 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 10:36:32,583 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.641e+02 1.959e+02 2.194e+02 3.203e+02, threshold=3.918e+02, percent-clipped=0.0 +2023-03-21 10:36:41,695 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116974.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:36:44,273 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116979.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:36:45,085 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-03-21 10:36:47,271 INFO [train.py:901] (1/2) Epoch 42, batch 1200, loss[loss=0.1218, simple_loss=0.2091, pruned_loss=0.01726, over 7296.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2098, pruned_loss=0.02345, over 1441014.24 frames. ], batch size: 80, lr: 3.93e-03, grad_scale: 8.0 +2023-03-21 10:37:03,317 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 10:37:10,010 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=117027.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:37:14,025 INFO [train.py:901] (1/2) Epoch 42, batch 1250, loss[loss=0.1379, simple_loss=0.2226, pruned_loss=0.02656, over 7241.00 frames. ], tot_loss[loss=0.1286, simple_loss=0.2102, pruned_loss=0.02351, over 1442122.07 frames. ], batch size: 93, lr: 3.93e-03, grad_scale: 8.0 +2023-03-21 10:37:16,084 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8601, 3.0049, 3.8177, 3.6785, 3.8284, 3.8277, 3.7689, 3.7284], + device='cuda:1'), covar=tensor([0.0032, 0.0133, 0.0030, 0.0036, 0.0033, 0.0031, 0.0049, 0.0047], + device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0071, 0.0058, 0.0058, 0.0056, 0.0060, 0.0049, 0.0078], + device='cuda:1'), out_proj_covar=tensor([8.2970e-05, 1.4241e-04, 1.0392e-04, 9.8061e-05, 9.4034e-05, 1.0437e-04, + 9.2983e-05, 1.4178e-04], device='cuda:1') +2023-03-21 10:37:24,586 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 1.700e+02 1.978e+02 2.314e+02 4.583e+02, threshold=3.956e+02, percent-clipped=1.0 +2023-03-21 10:37:26,671 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 10:37:26,724 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117060.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:37:27,241 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3481, 4.1389, 3.7342, 3.8502, 3.3621, 2.1368, 1.9616, 4.2778], + device='cuda:1'), covar=tensor([0.0041, 0.0066, 0.0108, 0.0073, 0.0141, 0.0631, 0.0666, 0.0050], + device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0095, 0.0114, 0.0097, 0.0133, 0.0137, 0.0130, 0.0107], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 10:37:30,554 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 10:37:31,568 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 10:37:39,144 INFO [train.py:901] (1/2) Epoch 42, batch 1300, loss[loss=0.1103, simple_loss=0.182, pruned_loss=0.01927, over 7041.00 frames. ], tot_loss[loss=0.1285, simple_loss=0.2101, pruned_loss=0.02345, over 1444525.60 frames. ], batch size: 35, lr: 3.93e-03, grad_scale: 8.0 +2023-03-21 10:37:48,328 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117102.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:37:56,035 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 10:37:58,676 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 10:38:02,271 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 10:38:02,452 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9334, 3.4224, 3.6105, 3.7032, 3.3852, 3.2249, 3.9409, 2.7638], + device='cuda:1'), covar=tensor([0.0434, 0.0564, 0.0713, 0.0666, 0.0780, 0.1017, 0.0747, 0.2278], + device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0336, 0.0271, 0.0354, 0.0285, 0.0286, 0.0344, 0.0242], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:38:05,800 INFO [train.py:901] (1/2) Epoch 42, batch 1350, loss[loss=0.1027, simple_loss=0.1809, pruned_loss=0.01231, over 7151.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2097, pruned_loss=0.0234, over 1444572.36 frames. ], batch size: 39, lr: 3.93e-03, grad_scale: 8.0 +2023-03-21 10:38:12,851 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 10:38:13,413 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=117150.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:38:16,415 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.654e+02 1.903e+02 2.260e+02 3.154e+02, threshold=3.806e+02, percent-clipped=0.0 +2023-03-21 10:38:17,589 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117158.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:38:19,109 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117161.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:38:29,493 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 +2023-03-21 10:38:31,676 INFO [train.py:901] (1/2) Epoch 42, batch 1400, loss[loss=0.1375, simple_loss=0.2138, pruned_loss=0.0306, over 7210.00 frames. ], tot_loss[loss=0.128, simple_loss=0.2093, pruned_loss=0.0234, over 1443618.27 frames. ], batch size: 50, lr: 3.93e-03, grad_scale: 8.0 +2023-03-21 10:38:44,768 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=117209.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:38:47,305 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 10:38:50,872 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117221.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:38:57,824 INFO [train.py:901] (1/2) Epoch 42, batch 1450, loss[loss=0.143, simple_loss=0.2236, pruned_loss=0.03121, over 7233.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2095, pruned_loss=0.02337, over 1442622.31 frames. ], batch size: 55, lr: 3.93e-03, grad_scale: 8.0 +2023-03-21 10:39:08,246 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.818e+02 2.119e+02 2.503e+02 5.217e+02, threshold=4.239e+02, percent-clipped=4.0 +2023-03-21 10:39:09,809 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 10:39:18,064 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117274.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:39:22,710 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117282.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:39:24,104 INFO [train.py:901] (1/2) Epoch 42, batch 1500, loss[loss=0.09888, simple_loss=0.165, pruned_loss=0.01638, over 6215.00 frames. ], tot_loss[loss=0.128, simple_loss=0.2093, pruned_loss=0.02336, over 1440634.70 frames. ], batch size: 26, lr: 3.93e-03, grad_scale: 8.0 +2023-03-21 10:39:27,150 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 10:39:41,327 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-03-21 10:39:42,601 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=117322.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:39:46,147 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117329.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:39:48,936 INFO [train.py:901] (1/2) Epoch 42, batch 1550, loss[loss=0.1322, simple_loss=0.2213, pruned_loss=0.02151, over 7319.00 frames. ], tot_loss[loss=0.128, simple_loss=0.2093, pruned_loss=0.0233, over 1441930.48 frames. ], batch size: 61, lr: 3.93e-03, grad_scale: 8.0 +2023-03-21 10:39:48,966 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 10:40:00,048 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.700e+02 1.983e+02 2.415e+02 3.192e+02, threshold=3.965e+02, percent-clipped=0.0 +2023-03-21 10:40:02,258 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117360.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:40:02,548 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 +2023-03-21 10:40:15,301 INFO [train.py:901] (1/2) Epoch 42, batch 1600, loss[loss=0.1301, simple_loss=0.2121, pruned_loss=0.02409, over 7327.00 frames. ], tot_loss[loss=0.1276, simple_loss=0.2089, pruned_loss=0.02316, over 1439039.84 frames. ], batch size: 44, lr: 3.92e-03, grad_scale: 8.0 +2023-03-21 10:40:17,909 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117390.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:40:20,366 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 10:40:20,879 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 10:40:24,279 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 10:40:24,941 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9912, 3.3761, 2.9830, 3.2148, 3.2504, 2.8554, 3.3624, 3.1100], + device='cuda:1'), covar=tensor([0.1183, 0.0602, 0.0668, 0.1290, 0.1176, 0.0689, 0.0530, 0.1228], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0060, 0.0069, 0.0061, 0.0058, 0.0064, 0.0057, 0.0055], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:40:27,424 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=117408.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:40:33,783 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 10:40:38,320 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 10:40:41,382 INFO [train.py:901] (1/2) Epoch 42, batch 1650, loss[loss=0.1248, simple_loss=0.2092, pruned_loss=0.02016, over 7213.00 frames. ], tot_loss[loss=0.1285, simple_loss=0.21, pruned_loss=0.0235, over 1440176.15 frames. ], batch size: 50, lr: 3.92e-03, grad_scale: 8.0 +2023-03-21 10:40:45,786 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 10:40:51,671 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.314e+02 1.848e+02 2.191e+02 2.608e+02 4.497e+02, threshold=4.382e+02, percent-clipped=1.0 +2023-03-21 10:40:53,416 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117458.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:41:03,889 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 10:41:06,954 INFO [train.py:901] (1/2) Epoch 42, batch 1700, loss[loss=0.1409, simple_loss=0.2208, pruned_loss=0.03051, over 7250.00 frames. ], tot_loss[loss=0.1288, simple_loss=0.2101, pruned_loss=0.02372, over 1442657.46 frames. ], batch size: 64, lr: 3.92e-03, grad_scale: 8.0 +2023-03-21 10:41:07,976 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 10:41:17,499 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=117506.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:41:19,008 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 10:41:32,724 INFO [train.py:901] (1/2) Epoch 42, batch 1750, loss[loss=0.1339, simple_loss=0.2112, pruned_loss=0.02831, over 7276.00 frames. ], tot_loss[loss=0.1287, simple_loss=0.2099, pruned_loss=0.02372, over 1440791.95 frames. ], batch size: 66, lr: 3.92e-03, grad_scale: 8.0 +2023-03-21 10:41:43,728 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.708e+02 1.965e+02 2.274e+02 3.472e+02, threshold=3.931e+02, percent-clipped=0.0 +2023-03-21 10:41:43,757 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 10:41:44,771 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 10:41:54,396 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117577.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:41:58,332 INFO [train.py:901] (1/2) Epoch 42, batch 1800, loss[loss=0.1268, simple_loss=0.2079, pruned_loss=0.02282, over 7327.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2093, pruned_loss=0.0236, over 1441784.86 frames. ], batch size: 54, lr: 3.92e-03, grad_scale: 8.0 +2023-03-21 10:42:06,734 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 10:42:18,952 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4279, 2.7144, 2.6080, 2.5645, 2.6040, 2.4607, 2.4884, 2.0378], + device='cuda:1'), covar=tensor([0.0507, 0.0355, 0.0378, 0.0197, 0.0645, 0.0596, 0.0325, 0.0384], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0040, 0.0041, 0.0040, 0.0038, 0.0038, 0.0044, 0.0043], + device='cuda:1'), out_proj_covar=tensor([1.0414e-04, 1.0399e-04, 1.0316e-04, 1.0136e-04, 9.9787e-05, 9.8862e-05, + 1.0890e-04, 1.0930e-04], device='cuda:1') +2023-03-21 10:42:20,330 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 10:42:24,842 INFO [train.py:901] (1/2) Epoch 42, batch 1850, loss[loss=0.1097, simple_loss=0.1894, pruned_loss=0.01498, over 7129.00 frames. ], tot_loss[loss=0.1288, simple_loss=0.2101, pruned_loss=0.02376, over 1441829.73 frames. ], batch size: 41, lr: 3.92e-03, grad_scale: 8.0 +2023-03-21 10:42:30,337 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 10:42:35,310 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.716e+02 2.093e+02 2.337e+02 4.292e+02, threshold=4.185e+02, percent-clipped=1.0 +2023-03-21 10:42:37,010 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2092, 3.0092, 2.0110, 3.5260, 3.6783, 3.5904, 3.2709, 3.2730], + device='cuda:1'), covar=tensor([0.2477, 0.0863, 0.4436, 0.0464, 0.0241, 0.0216, 0.0332, 0.0324], + device='cuda:1'), in_proj_covar=tensor([0.0255, 0.0234, 0.0249, 0.0261, 0.0201, 0.0206, 0.0222, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 10:42:45,457 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 10:42:49,990 INFO [train.py:901] (1/2) Epoch 42, batch 1900, loss[loss=0.108, simple_loss=0.1803, pruned_loss=0.0179, over 6969.00 frames. ], tot_loss[loss=0.1285, simple_loss=0.2099, pruned_loss=0.02361, over 1442776.96 frames. ], batch size: 35, lr: 3.92e-03, grad_scale: 8.0 +2023-03-21 10:42:50,075 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117685.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:43:10,322 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 10:43:16,250 INFO [train.py:901] (1/2) Epoch 42, batch 1950, loss[loss=0.1161, simple_loss=0.198, pruned_loss=0.0171, over 7117.00 frames. ], tot_loss[loss=0.1287, simple_loss=0.2103, pruned_loss=0.02356, over 1443705.65 frames. ], batch size: 41, lr: 3.92e-03, grad_scale: 8.0 +2023-03-21 10:43:21,194 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 10:43:26,198 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 10:43:26,668 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 1.776e+02 2.090e+02 2.427e+02 5.955e+02, threshold=4.180e+02, percent-clipped=1.0 +2023-03-21 10:43:26,708 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 10:43:41,890 INFO [train.py:901] (1/2) Epoch 42, batch 2000, loss[loss=0.1489, simple_loss=0.2154, pruned_loss=0.04119, over 7316.00 frames. ], tot_loss[loss=0.1288, simple_loss=0.2102, pruned_loss=0.02374, over 1444861.20 frames. ], batch size: 49, lr: 3.92e-03, grad_scale: 8.0 +2023-03-21 10:43:43,958 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 10:43:50,590 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7557, 3.9124, 3.7208, 3.8887, 3.7332, 3.8338, 4.1695, 4.1960], + device='cuda:1'), covar=tensor([0.0239, 0.0177, 0.0249, 0.0193, 0.0286, 0.0416, 0.0243, 0.0199], + device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0127, 0.0121, 0.0126, 0.0114, 0.0102, 0.0098, 0.0102], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 10:43:56,182 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 10:44:04,178 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 10:44:08,066 INFO [train.py:901] (1/2) Epoch 42, batch 2050, loss[loss=0.1453, simple_loss=0.2143, pruned_loss=0.03816, over 7221.00 frames. ], tot_loss[loss=0.1292, simple_loss=0.2108, pruned_loss=0.02384, over 1445035.36 frames. ], batch size: 45, lr: 3.92e-03, grad_scale: 8.0 +2023-03-21 10:44:18,463 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.201e+02 1.814e+02 2.172e+02 2.477e+02 4.017e+02, threshold=4.343e+02, percent-clipped=0.0 +2023-03-21 10:44:29,770 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4272, 4.0282, 4.1000, 4.0587, 4.0617, 3.9333, 4.3089, 3.7121], + device='cuda:1'), covar=tensor([0.0157, 0.0181, 0.0130, 0.0196, 0.0430, 0.0146, 0.0145, 0.0215], + device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0105, 0.0105, 0.0090, 0.0180, 0.0110, 0.0109, 0.0116], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 10:44:29,787 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117877.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:44:30,256 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3600, 4.9164, 4.9810, 4.9046, 4.8140, 4.5161, 5.0020, 4.8373], + device='cuda:1'), covar=tensor([0.0495, 0.0411, 0.0383, 0.0505, 0.0333, 0.0403, 0.0311, 0.0392], + device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0265, 0.0208, 0.0207, 0.0161, 0.0235, 0.0216, 0.0149], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:44:33,671 INFO [train.py:901] (1/2) Epoch 42, batch 2100, loss[loss=0.0957, simple_loss=0.1595, pruned_loss=0.01594, over 6051.00 frames. ], tot_loss[loss=0.1292, simple_loss=0.2107, pruned_loss=0.02383, over 1443818.06 frames. ], batch size: 25, lr: 3.92e-03, grad_scale: 8.0 +2023-03-21 10:44:38,089 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 10:44:40,679 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 10:44:51,799 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9836, 2.7288, 3.2908, 2.9135, 3.2218, 3.0448, 2.7274, 3.1393], + device='cuda:1'), covar=tensor([0.1350, 0.0814, 0.0996, 0.1464, 0.0827, 0.0936, 0.1962, 0.1742], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0070, 0.0053, 0.0053, 0.0053, 0.0052, 0.0071, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 10:44:52,871 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4830, 2.9962, 3.3843, 3.3982, 3.1431, 2.9310, 3.6134, 2.4471], + device='cuda:1'), covar=tensor([0.0504, 0.0542, 0.0741, 0.0680, 0.0661, 0.0933, 0.0587, 0.2569], + device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0337, 0.0268, 0.0352, 0.0281, 0.0285, 0.0344, 0.0242], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:44:54,188 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=117925.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:44:59,105 INFO [train.py:901] (1/2) Epoch 42, batch 2150, loss[loss=0.1294, simple_loss=0.209, pruned_loss=0.02492, over 7270.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.211, pruned_loss=0.02408, over 1441358.38 frames. ], batch size: 52, lr: 3.92e-03, grad_scale: 8.0 +2023-03-21 10:45:10,028 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.728e+02 2.032e+02 2.462e+02 4.551e+02, threshold=4.063e+02, percent-clipped=1.0 +2023-03-21 10:45:25,245 INFO [train.py:901] (1/2) Epoch 42, batch 2200, loss[loss=0.1426, simple_loss=0.2251, pruned_loss=0.03, over 6763.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.2113, pruned_loss=0.02395, over 1442715.91 frames. ], batch size: 106, lr: 3.91e-03, grad_scale: 8.0 +2023-03-21 10:45:25,260 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 10:45:25,340 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117985.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:45:27,942 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7373, 3.2321, 3.5304, 3.5822, 3.2915, 3.1201, 3.8658, 2.6765], + device='cuda:1'), covar=tensor([0.0451, 0.0576, 0.0705, 0.0667, 0.0835, 0.1012, 0.0750, 0.2443], + device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0334, 0.0266, 0.0350, 0.0279, 0.0283, 0.0341, 0.0240], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:45:50,276 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=118033.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:45:51,235 INFO [train.py:901] (1/2) Epoch 42, batch 2250, loss[loss=0.1402, simple_loss=0.2247, pruned_loss=0.02783, over 7307.00 frames. ], tot_loss[loss=0.1294, simple_loss=0.2112, pruned_loss=0.02383, over 1442871.07 frames. ], batch size: 83, lr: 3.91e-03, grad_scale: 8.0 +2023-03-21 10:45:53,890 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118040.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:45:59,344 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 10:45:59,826 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 10:46:01,750 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.280e+02 1.865e+02 2.098e+02 2.420e+02 4.681e+02, threshold=4.196e+02, percent-clipped=2.0 +2023-03-21 10:46:13,110 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 10:46:17,176 INFO [train.py:901] (1/2) Epoch 42, batch 2300, loss[loss=0.1597, simple_loss=0.2396, pruned_loss=0.03992, over 6571.00 frames. ], tot_loss[loss=0.1302, simple_loss=0.2118, pruned_loss=0.02426, over 1443796.89 frames. ], batch size: 106, lr: 3.91e-03, grad_scale: 8.0 +2023-03-21 10:46:25,273 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118101.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:46:42,780 INFO [train.py:901] (1/2) Epoch 42, batch 2350, loss[loss=0.1188, simple_loss=0.2067, pruned_loss=0.01547, over 7359.00 frames. ], tot_loss[loss=0.1301, simple_loss=0.2117, pruned_loss=0.02423, over 1443462.67 frames. ], batch size: 73, lr: 3.91e-03, grad_scale: 8.0 +2023-03-21 10:46:53,311 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.246e+02 1.681e+02 1.961e+02 2.398e+02 5.180e+02, threshold=3.922e+02, percent-clipped=2.0 +2023-03-21 10:46:59,501 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 10:47:00,067 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118168.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:47:06,088 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 10:47:08,469 INFO [train.py:901] (1/2) Epoch 42, batch 2400, loss[loss=0.1359, simple_loss=0.21, pruned_loss=0.0309, over 7328.00 frames. ], tot_loss[loss=0.1294, simple_loss=0.211, pruned_loss=0.02394, over 1441922.23 frames. ], batch size: 44, lr: 3.91e-03, grad_scale: 8.0 +2023-03-21 10:47:17,338 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 10:47:20,945 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 10:47:31,500 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118229.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 10:47:34,326 INFO [train.py:901] (1/2) Epoch 42, batch 2450, loss[loss=0.1267, simple_loss=0.2075, pruned_loss=0.02296, over 7290.00 frames. ], tot_loss[loss=0.1291, simple_loss=0.2106, pruned_loss=0.0238, over 1444466.36 frames. ], batch size: 57, lr: 3.91e-03, grad_scale: 8.0 +2023-03-21 10:47:36,407 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7377, 2.3828, 2.8844, 2.7257, 2.9762, 2.7694, 2.4778, 3.0246], + device='cuda:1'), covar=tensor([0.1329, 0.0781, 0.0846, 0.1235, 0.0781, 0.0885, 0.1698, 0.0889], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0070, 0.0053, 0.0052, 0.0052, 0.0051, 0.0070, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 10:47:45,320 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.245e+02 1.774e+02 2.090e+02 2.469e+02 5.091e+02, threshold=4.180e+02, percent-clipped=0.0 +2023-03-21 10:47:47,400 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 10:47:52,044 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118269.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:47:59,878 INFO [train.py:901] (1/2) Epoch 42, batch 2500, loss[loss=0.137, simple_loss=0.228, pruned_loss=0.02303, over 7148.00 frames. ], tot_loss[loss=0.1294, simple_loss=0.2111, pruned_loss=0.02383, over 1446042.33 frames. ], batch size: 98, lr: 3.91e-03, grad_scale: 16.0 +2023-03-21 10:48:02,680 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118289.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:48:06,734 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4333, 1.7861, 1.3741, 1.7400, 1.8441, 1.7411, 1.6641, 1.4180], + device='cuda:1'), covar=tensor([0.0252, 0.0319, 0.0533, 0.0270, 0.0203, 0.0165, 0.0201, 0.0264], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0037, 0.0038, 0.0037, 0.0035, 0.0039, 0.0047], + device='cuda:1'), out_proj_covar=tensor([4.4468e-05, 4.3891e-05, 4.2239e-05, 4.2493e-05, 4.1096e-05, 3.9382e-05, + 4.3498e-05, 5.1896e-05], device='cuda:1') +2023-03-21 10:48:12,612 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 10:48:22,730 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118329.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:48:23,262 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118330.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:48:25,608 INFO [train.py:901] (1/2) Epoch 42, batch 2550, loss[loss=0.1361, simple_loss=0.2137, pruned_loss=0.02926, over 7324.00 frames. ], tot_loss[loss=0.1294, simple_loss=0.2109, pruned_loss=0.02392, over 1445401.41 frames. ], batch size: 59, lr: 3.91e-03, grad_scale: 8.0 +2023-03-21 10:48:33,833 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118350.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:48:37,184 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.126e+02 1.748e+02 2.085e+02 2.434e+02 8.128e+02, threshold=4.170e+02, percent-clipped=2.0 +2023-03-21 10:48:51,753 INFO [train.py:901] (1/2) Epoch 42, batch 2600, loss[loss=0.1255, simple_loss=0.2096, pruned_loss=0.02073, over 7329.00 frames. ], tot_loss[loss=0.1289, simple_loss=0.2105, pruned_loss=0.02363, over 1444706.97 frames. ], batch size: 83, lr: 3.91e-03, grad_scale: 8.0 +2023-03-21 10:48:54,435 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118390.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:48:57,317 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118396.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:49:16,727 INFO [train.py:901] (1/2) Epoch 42, batch 2650, loss[loss=0.1443, simple_loss=0.2266, pruned_loss=0.03102, over 7224.00 frames. ], tot_loss[loss=0.1292, simple_loss=0.211, pruned_loss=0.02367, over 1444140.67 frames. ], batch size: 93, lr: 3.91e-03, grad_scale: 8.0 +2023-03-21 10:49:27,697 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.174e+02 1.667e+02 1.916e+02 2.431e+02 3.839e+02, threshold=3.833e+02, percent-clipped=0.0 +2023-03-21 10:49:33,608 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9806, 3.2966, 2.7755, 3.1448, 3.2405, 2.6114, 3.1662, 2.9490], + device='cuda:1'), covar=tensor([0.0655, 0.0682, 0.1705, 0.0957, 0.0934, 0.0990, 0.1044, 0.1141], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0060, 0.0069, 0.0061, 0.0058, 0.0064, 0.0057, 0.0055], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:49:35,701 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.75 vs. limit=5.0 +2023-03-21 10:49:39,474 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9089, 2.6924, 2.8792, 2.9752, 2.5965, 2.6311, 3.0795, 2.1780], + device='cuda:1'), covar=tensor([0.0650, 0.0761, 0.0792, 0.0797, 0.0662, 0.1054, 0.0849, 0.2577], + device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0336, 0.0268, 0.0352, 0.0281, 0.0285, 0.0344, 0.0242], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:49:41,249 INFO [train.py:901] (1/2) Epoch 42, batch 2700, loss[loss=0.1331, simple_loss=0.2069, pruned_loss=0.02965, over 7273.00 frames. ], tot_loss[loss=0.1289, simple_loss=0.2106, pruned_loss=0.02358, over 1442592.11 frames. ], batch size: 47, lr: 3.91e-03, grad_scale: 8.0 +2023-03-21 10:50:00,191 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118524.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 10:50:05,544 INFO [train.py:901] (1/2) Epoch 42, batch 2750, loss[loss=0.1322, simple_loss=0.2154, pruned_loss=0.02451, over 7319.00 frames. ], tot_loss[loss=0.1296, simple_loss=0.2111, pruned_loss=0.024, over 1444073.31 frames. ], batch size: 80, lr: 3.91e-03, grad_scale: 8.0 +2023-03-21 10:50:16,537 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.177e+02 1.730e+02 2.063e+02 2.457e+02 5.328e+02, threshold=4.125e+02, percent-clipped=3.0 +2023-03-21 10:50:30,176 INFO [train.py:901] (1/2) Epoch 42, batch 2800, loss[loss=0.1394, simple_loss=0.2155, pruned_loss=0.03168, over 7299.00 frames. ], tot_loss[loss=0.1298, simple_loss=0.2113, pruned_loss=0.02415, over 1443584.78 frames. ], batch size: 70, lr: 3.91e-03, grad_scale: 8.0 +2023-03-21 10:50:53,487 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 10:50:53,539 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 10:50:53,564 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 10:50:53,581 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 10:50:54,052 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 10:50:54,072 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 10:50:54,082 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 10:50:54,178 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 10:50:54,250 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 10:50:54,266 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 10:50:54,281 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 10:50:54,445 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 10:50:54,450 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 10:50:54,457 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 10:50:54,599 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 10:50:54,602 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 10:50:54,979 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 10:50:58,604 INFO [train.py:901] (1/2) Epoch 43, batch 0, loss[loss=0.1167, simple_loss=0.2097, pruned_loss=0.01183, over 7243.00 frames. ], tot_loss[loss=0.1167, simple_loss=0.2097, pruned_loss=0.01183, over 7243.00 frames. ], batch size: 89, lr: 3.86e-03, grad_scale: 8.0 +2023-03-21 10:50:58,605 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 10:51:03,339 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3216, 3.1050, 3.6939, 3.1812, 3.6103, 3.4228, 3.0950, 3.4134], + device='cuda:1'), covar=tensor([0.0913, 0.0513, 0.0527, 0.1204, 0.0410, 0.0509, 0.1269, 0.0947], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0070, 0.0053, 0.0052, 0.0052, 0.0050, 0.0070, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 10:51:24,839 INFO [train.py:935] (1/2) Epoch 43, validation: loss=0.1646, simple_loss=0.2571, pruned_loss=0.03606, over 1622729.00 frames. +2023-03-21 10:51:24,840 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 10:51:30,887 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 10:51:33,558 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118625.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:51:36,730 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118631.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:51:37,748 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4193, 2.0981, 2.2140, 3.5284, 1.8112, 3.4557, 1.3189, 3.2554], + device='cuda:1'), covar=tensor([0.0241, 0.1696, 0.2016, 0.0286, 0.4426, 0.0315, 0.1569, 0.0410], + device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0246, 0.0257, 0.0212, 0.0249, 0.0219, 0.0222, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 10:51:42,071 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 10:51:43,658 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118645.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:51:48,659 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 10:51:49,633 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.053e+02 1.693e+02 1.954e+02 2.272e+02 3.982e+02, threshold=3.907e+02, percent-clipped=0.0 +2023-03-21 10:51:50,635 INFO [train.py:901] (1/2) Epoch 43, batch 50, loss[loss=0.1342, simple_loss=0.2184, pruned_loss=0.02495, over 7333.00 frames. ], tot_loss[loss=0.1293, simple_loss=0.2109, pruned_loss=0.02389, over 327228.99 frames. ], batch size: 75, lr: 3.86e-03, grad_scale: 8.0 +2023-03-21 10:51:51,146 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 10:51:54,219 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 10:52:04,427 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118685.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:52:07,996 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118692.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:52:09,870 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118696.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:52:11,753 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 10:52:12,212 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 10:52:16,188 INFO [train.py:901] (1/2) Epoch 43, batch 100, loss[loss=0.1299, simple_loss=0.2126, pruned_loss=0.02361, over 7269.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.209, pruned_loss=0.02357, over 572157.86 frames. ], batch size: 89, lr: 3.86e-03, grad_scale: 8.0 +2023-03-21 10:52:34,363 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=118744.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:52:40,763 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.242e+02 1.769e+02 1.999e+02 2.347e+02 3.939e+02, threshold=3.999e+02, percent-clipped=2.0 +2023-03-21 10:52:41,760 INFO [train.py:901] (1/2) Epoch 43, batch 150, loss[loss=0.1268, simple_loss=0.216, pruned_loss=0.0188, over 7139.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2091, pruned_loss=0.02365, over 764971.05 frames. ], batch size: 98, lr: 3.86e-03, grad_scale: 8.0 +2023-03-21 10:53:00,662 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.32 vs. limit=5.0 +2023-03-21 10:53:04,025 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3721, 2.9543, 2.2777, 3.1552, 3.0497, 3.3731, 2.7979, 2.7995], + device='cuda:1'), covar=tensor([0.2391, 0.1027, 0.3851, 0.0906, 0.0281, 0.0383, 0.0406, 0.0429], + device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0231, 0.0244, 0.0258, 0.0199, 0.0203, 0.0218, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 10:53:08,472 INFO [train.py:901] (1/2) Epoch 43, batch 200, loss[loss=0.1261, simple_loss=0.2102, pruned_loss=0.02105, over 7278.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2095, pruned_loss=0.02347, over 915173.05 frames. ], batch size: 68, lr: 3.86e-03, grad_scale: 8.0 +2023-03-21 10:53:13,183 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 10:53:16,329 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118824.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:53:18,254 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 10:53:23,756 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 10:53:32,655 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.321e+02 1.696e+02 2.045e+02 2.404e+02 3.575e+02, threshold=4.091e+02, percent-clipped=0.0 +2023-03-21 10:53:33,660 INFO [train.py:901] (1/2) Epoch 43, batch 250, loss[loss=0.1259, simple_loss=0.2038, pruned_loss=0.02402, over 7228.00 frames. ], tot_loss[loss=0.1287, simple_loss=0.2101, pruned_loss=0.02364, over 1033716.98 frames. ], batch size: 45, lr: 3.85e-03, grad_scale: 8.0 +2023-03-21 10:53:36,837 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 10:53:40,853 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=118872.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:53:44,120 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-03-21 10:53:58,100 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 10:54:00,030 INFO [train.py:901] (1/2) Epoch 43, batch 300, loss[loss=0.1316, simple_loss=0.2112, pruned_loss=0.02599, over 7291.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2089, pruned_loss=0.02362, over 1124061.20 frames. ], batch size: 68, lr: 3.85e-03, grad_scale: 8.0 +2023-03-21 10:54:06,605 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 10:54:08,119 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118925.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:54:18,096 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118945.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:54:24,621 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.093e+02 1.683e+02 1.946e+02 2.383e+02 3.462e+02, threshold=3.893e+02, percent-clipped=0.0 +2023-03-21 10:54:25,627 INFO [train.py:901] (1/2) Epoch 43, batch 350, loss[loss=0.109, simple_loss=0.1899, pruned_loss=0.01404, over 7153.00 frames. ], tot_loss[loss=0.128, simple_loss=0.209, pruned_loss=0.02349, over 1196718.20 frames. ], batch size: 41, lr: 3.85e-03, grad_scale: 8.0 +2023-03-21 10:54:33,248 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=118973.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:54:39,259 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118985.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:54:40,218 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118987.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:54:42,179 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 10:54:43,191 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=118993.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:54:51,698 INFO [train.py:901] (1/2) Epoch 43, batch 400, loss[loss=0.1164, simple_loss=0.1984, pruned_loss=0.01718, over 7283.00 frames. ], tot_loss[loss=0.1276, simple_loss=0.2085, pruned_loss=0.02339, over 1250235.07 frames. ], batch size: 57, lr: 3.85e-03, grad_scale: 8.0 +2023-03-21 10:55:03,877 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=119033.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:55:16,237 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.313e+02 1.703e+02 1.963e+02 2.470e+02 3.924e+02, threshold=3.926e+02, percent-clipped=1.0 +2023-03-21 10:55:17,235 INFO [train.py:901] (1/2) Epoch 43, batch 450, loss[loss=0.1241, simple_loss=0.2079, pruned_loss=0.02012, over 7337.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2093, pruned_loss=0.02377, over 1292310.83 frames. ], batch size: 75, lr: 3.85e-03, grad_scale: 8.0 +2023-03-21 10:55:23,779 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 10:55:24,278 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 10:55:25,371 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8380, 3.1738, 2.8594, 3.0536, 3.1639, 2.5732, 3.0377, 2.9004], + device='cuda:1'), covar=tensor([0.0763, 0.0837, 0.0988, 0.0992, 0.0811, 0.0971, 0.1072, 0.0822], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0060, 0.0069, 0.0061, 0.0057, 0.0064, 0.0057, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:55:25,876 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119075.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:55:34,853 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119093.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:55:42,757 INFO [train.py:901] (1/2) Epoch 43, batch 500, loss[loss=0.1311, simple_loss=0.2118, pruned_loss=0.0252, over 7299.00 frames. ], tot_loss[loss=0.1285, simple_loss=0.2093, pruned_loss=0.02384, over 1325557.79 frames. ], batch size: 86, lr: 3.85e-03, grad_scale: 8.0 +2023-03-21 10:55:57,080 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119136.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:55:57,933 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 10:55:59,452 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 10:55:59,929 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 10:56:02,413 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 10:56:06,653 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119154.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 10:56:07,514 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 10:56:08,019 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.252e+02 1.731e+02 2.047e+02 2.533e+02 3.684e+02, threshold=4.094e+02, percent-clipped=0.0 +2023-03-21 10:56:09,059 INFO [train.py:901] (1/2) Epoch 43, batch 550, loss[loss=0.1208, simple_loss=0.1972, pruned_loss=0.02214, over 7325.00 frames. ], tot_loss[loss=0.128, simple_loss=0.2088, pruned_loss=0.02363, over 1351855.15 frames. ], batch size: 44, lr: 3.85e-03, grad_scale: 8.0 +2023-03-21 10:56:18,036 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 10:56:25,945 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 10:56:26,616 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0526, 3.6405, 3.7671, 3.7696, 3.5748, 3.2674, 3.9196, 2.8219], + device='cuda:1'), covar=tensor([0.0781, 0.0505, 0.0631, 0.0744, 0.0805, 0.1006, 0.0639, 0.2505], + device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0341, 0.0273, 0.0357, 0.0285, 0.0289, 0.0347, 0.0244], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:56:29,453 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 10:56:34,167 INFO [train.py:901] (1/2) Epoch 43, batch 600, loss[loss=0.125, simple_loss=0.1978, pruned_loss=0.02607, over 7271.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.2089, pruned_loss=0.02331, over 1372285.62 frames. ], batch size: 52, lr: 3.85e-03, grad_scale: 8.0 +2023-03-21 10:56:36,832 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 10:56:53,677 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 10:56:59,203 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.206e+02 1.745e+02 2.072e+02 2.457e+02 4.750e+02, threshold=4.144e+02, percent-clipped=3.0 +2023-03-21 10:57:00,224 INFO [train.py:901] (1/2) Epoch 43, batch 650, loss[loss=0.1395, simple_loss=0.2122, pruned_loss=0.0334, over 7243.00 frames. ], tot_loss[loss=0.1274, simple_loss=0.2086, pruned_loss=0.02317, over 1387159.91 frames. ], batch size: 45, lr: 3.85e-03, grad_scale: 8.0 +2023-03-21 10:57:02,302 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 10:57:14,740 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119287.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:57:19,273 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 10:57:26,472 INFO [train.py:901] (1/2) Epoch 43, batch 700, loss[loss=0.1249, simple_loss=0.2054, pruned_loss=0.02223, over 7280.00 frames. ], tot_loss[loss=0.1273, simple_loss=0.2085, pruned_loss=0.02306, over 1398737.54 frames. ], batch size: 70, lr: 3.85e-03, grad_scale: 4.0 +2023-03-21 10:57:29,456 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 10:57:40,127 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=119335.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:57:50,793 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-03-21 10:57:51,953 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.591e+02 1.956e+02 2.341e+02 4.303e+02, threshold=3.912e+02, percent-clipped=1.0 +2023-03-21 10:57:51,972 INFO [train.py:901] (1/2) Epoch 43, batch 750, loss[loss=0.1283, simple_loss=0.2147, pruned_loss=0.02097, over 7274.00 frames. ], tot_loss[loss=0.128, simple_loss=0.2093, pruned_loss=0.0233, over 1408124.58 frames. ], batch size: 64, lr: 3.85e-03, grad_scale: 2.0 +2023-03-21 10:57:52,956 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 10:57:53,478 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 10:58:08,067 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 10:58:13,155 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 10:58:17,911 INFO [train.py:901] (1/2) Epoch 43, batch 800, loss[loss=0.1131, simple_loss=0.1946, pruned_loss=0.01576, over 7135.00 frames. ], tot_loss[loss=0.1283, simple_loss=0.2096, pruned_loss=0.0235, over 1417857.27 frames. ], batch size: 41, lr: 3.85e-03, grad_scale: 4.0 +2023-03-21 10:58:19,990 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 10:58:21,001 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 10:58:25,836 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-03-21 10:58:29,563 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119431.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:58:30,118 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9916, 1.6690, 2.3072, 2.4800, 2.3255, 2.5366, 2.4123, 2.6248], + device='cuda:1'), covar=tensor([0.2190, 0.4346, 0.2558, 0.2625, 0.2660, 0.3709, 0.2449, 0.1807], + device='cuda:1'), in_proj_covar=tensor([0.0080, 0.0082, 0.0075, 0.0068, 0.0067, 0.0066, 0.0107, 0.0070], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:58:32,492 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 10:58:38,679 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119449.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 10:58:39,724 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6362, 2.3022, 2.7320, 2.7681, 2.8764, 2.7652, 2.3995, 2.7388], + device='cuda:1'), covar=tensor([0.1750, 0.0881, 0.1346, 0.0787, 0.0757, 0.0745, 0.1493, 0.1085], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0070, 0.0053, 0.0052, 0.0052, 0.0051, 0.0070, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 10:58:43,522 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 1.777e+02 2.026e+02 2.373e+02 4.143e+02, threshold=4.052e+02, percent-clipped=2.0 +2023-03-21 10:58:43,541 INFO [train.py:901] (1/2) Epoch 43, batch 850, loss[loss=0.1264, simple_loss=0.211, pruned_loss=0.02085, over 7247.00 frames. ], tot_loss[loss=0.1288, simple_loss=0.21, pruned_loss=0.02384, over 1423313.48 frames. ], batch size: 55, lr: 3.85e-03, grad_scale: 4.0 +2023-03-21 10:58:44,233 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8423, 3.2839, 3.5657, 3.5597, 3.3698, 3.0834, 3.8060, 2.6892], + device='cuda:1'), covar=tensor([0.0575, 0.0592, 0.0713, 0.0686, 0.0702, 0.1019, 0.0620, 0.2360], + device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0340, 0.0273, 0.0358, 0.0284, 0.0290, 0.0348, 0.0245], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:58:51,170 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 10:58:51,180 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 10:58:53,768 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0090, 3.3272, 2.8660, 3.1348, 3.1844, 2.8167, 3.3164, 2.9627], + device='cuda:1'), covar=tensor([0.1122, 0.0696, 0.0691, 0.1149, 0.1261, 0.1055, 0.0690, 0.1322], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0060, 0.0069, 0.0061, 0.0057, 0.0064, 0.0057, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:58:56,216 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 10:58:57,953 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-03-21 10:59:00,236 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 10:59:09,900 INFO [train.py:901] (1/2) Epoch 43, batch 900, loss[loss=0.1181, simple_loss=0.199, pruned_loss=0.01864, over 7350.00 frames. ], tot_loss[loss=0.1285, simple_loss=0.2094, pruned_loss=0.02375, over 1427604.11 frames. ], batch size: 44, lr: 3.84e-03, grad_scale: 4.0 +2023-03-21 10:59:12,495 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119514.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:59:30,494 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0456, 2.7599, 3.0301, 2.9648, 2.7399, 2.6444, 3.2114, 2.2729], + device='cuda:1'), covar=tensor([0.0452, 0.0571, 0.0720, 0.0618, 0.0680, 0.0966, 0.0702, 0.2419], + device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0337, 0.0271, 0.0354, 0.0282, 0.0287, 0.0345, 0.0243], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 10:59:34,900 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.200e+02 1.779e+02 2.084e+02 2.521e+02 3.784e+02, threshold=4.168e+02, percent-clipped=0.0 +2023-03-21 10:59:34,919 INFO [train.py:901] (1/2) Epoch 43, batch 950, loss[loss=0.1443, simple_loss=0.228, pruned_loss=0.03032, over 7237.00 frames. ], tot_loss[loss=0.1278, simple_loss=0.209, pruned_loss=0.02336, over 1432541.44 frames. ], batch size: 64, lr: 3.84e-03, grad_scale: 4.0 +2023-03-21 10:59:38,008 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 10:59:43,642 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119575.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 10:59:46,163 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8706, 2.3988, 2.9685, 2.7810, 3.0820, 2.9073, 2.5321, 2.9067], + device='cuda:1'), covar=tensor([0.1136, 0.0699, 0.1156, 0.1245, 0.0648, 0.0783, 0.1967, 0.1173], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0070, 0.0053, 0.0052, 0.0052, 0.0051, 0.0071, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 10:59:59,158 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3214, 4.0051, 4.0739, 4.0309, 3.7674, 3.9027, 4.1866, 3.7414], + device='cuda:1'), covar=tensor([0.0181, 0.0171, 0.0119, 0.0200, 0.0579, 0.0150, 0.0179, 0.0191], + device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0105, 0.0105, 0.0091, 0.0183, 0.0110, 0.0109, 0.0117], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 11:00:01,534 INFO [train.py:901] (1/2) Epoch 43, batch 1000, loss[loss=0.1438, simple_loss=0.2224, pruned_loss=0.0326, over 7250.00 frames. ], tot_loss[loss=0.1283, simple_loss=0.2095, pruned_loss=0.02355, over 1435278.52 frames. ], batch size: 64, lr: 3.84e-03, grad_scale: 4.0 +2023-03-21 11:00:01,546 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 11:00:20,420 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 11:00:27,113 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.326e+02 1.691e+02 2.007e+02 2.275e+02 7.458e+02, threshold=4.015e+02, percent-clipped=1.0 +2023-03-21 11:00:27,132 INFO [train.py:901] (1/2) Epoch 43, batch 1050, loss[loss=0.129, simple_loss=0.2122, pruned_loss=0.02292, over 7259.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2099, pruned_loss=0.02347, over 1437913.51 frames. ], batch size: 89, lr: 3.84e-03, grad_scale: 4.0 +2023-03-21 11:00:30,293 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119665.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:00:44,012 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 11:00:48,550 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 11:00:52,937 INFO [train.py:901] (1/2) Epoch 43, batch 1100, loss[loss=0.1325, simple_loss=0.2133, pruned_loss=0.02579, over 7102.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2096, pruned_loss=0.02345, over 1438217.73 frames. ], batch size: 98, lr: 3.84e-03, grad_scale: 4.0 +2023-03-21 11:00:54,143 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-03-21 11:01:01,433 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119726.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:01:03,840 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119731.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:01:13,563 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119749.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 11:01:17,994 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 11:01:18,505 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 11:01:18,974 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.320e+02 1.743e+02 2.054e+02 2.529e+02 4.457e+02, threshold=4.108e+02, percent-clipped=1.0 +2023-03-21 11:01:18,992 INFO [train.py:901] (1/2) Epoch 43, batch 1150, loss[loss=0.1598, simple_loss=0.2317, pruned_loss=0.04392, over 7240.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2092, pruned_loss=0.02355, over 1436400.96 frames. ], batch size: 55, lr: 3.84e-03, grad_scale: 4.0 +2023-03-21 11:01:28,899 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=119779.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:01:29,349 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 11:01:29,866 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 11:01:37,927 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=119797.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:01:44,226 INFO [train.py:901] (1/2) Epoch 43, batch 1200, loss[loss=0.1272, simple_loss=0.2119, pruned_loss=0.02129, over 7342.00 frames. ], tot_loss[loss=0.1283, simple_loss=0.2095, pruned_loss=0.02359, over 1437227.57 frames. ], batch size: 63, lr: 3.84e-03, grad_scale: 8.0 +2023-03-21 11:02:03,037 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 11:02:03,634 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2291, 2.4930, 2.7115, 2.2027, 2.3539, 2.3010, 2.1732, 1.8251], + device='cuda:1'), covar=tensor([0.0530, 0.0369, 0.0219, 0.0362, 0.0437, 0.0418, 0.0351, 0.0418], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0040, 0.0040, 0.0040, 0.0037, 0.0037, 0.0043, 0.0043], + device='cuda:1'), out_proj_covar=tensor([1.0326e-04, 1.0258e-04, 1.0145e-04, 1.0121e-04, 9.8119e-05, 9.7465e-05, + 1.0754e-04, 1.0797e-04], device='cuda:1') +2023-03-21 11:02:06,069 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8120, 3.2359, 2.8554, 2.9622, 3.1386, 2.7496, 3.0332, 2.9196], + device='cuda:1'), covar=tensor([0.0810, 0.0540, 0.0779, 0.1093, 0.0944, 0.0621, 0.0679, 0.0953], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0060, 0.0070, 0.0061, 0.0058, 0.0065, 0.0058, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:02:10,429 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.203e+02 1.784e+02 2.054e+02 2.428e+02 4.631e+02, threshold=4.108e+02, percent-clipped=3.0 +2023-03-21 11:02:10,447 INFO [train.py:901] (1/2) Epoch 43, batch 1250, loss[loss=0.1041, simple_loss=0.1662, pruned_loss=0.02099, over 6087.00 frames. ], tot_loss[loss=0.1283, simple_loss=0.2096, pruned_loss=0.0235, over 1435985.70 frames. ], batch size: 26, lr: 3.84e-03, grad_scale: 8.0 +2023-03-21 11:02:16,009 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119870.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:02:25,919 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 11:02:26,077 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4748, 2.9591, 3.2683, 3.3318, 2.9274, 2.8515, 3.3979, 2.3583], + device='cuda:1'), covar=tensor([0.0470, 0.0498, 0.0878, 0.0741, 0.0676, 0.1047, 0.0741, 0.2855], + device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0337, 0.0272, 0.0354, 0.0283, 0.0288, 0.0345, 0.0242], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:02:30,412 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 11:02:31,432 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 11:02:35,439 INFO [train.py:901] (1/2) Epoch 43, batch 1300, loss[loss=0.1313, simple_loss=0.2119, pruned_loss=0.02541, over 7233.00 frames. ], tot_loss[loss=0.129, simple_loss=0.2104, pruned_loss=0.02376, over 1440070.90 frames. ], batch size: 45, lr: 3.84e-03, grad_scale: 8.0 +2023-03-21 11:02:55,501 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-03-21 11:02:55,659 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 11:02:58,125 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 11:03:01,609 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.277e+02 1.695e+02 1.928e+02 2.177e+02 3.614e+02, threshold=3.856e+02, percent-clipped=0.0 +2023-03-21 11:03:01,627 INFO [train.py:901] (1/2) Epoch 43, batch 1350, loss[loss=0.1414, simple_loss=0.2315, pruned_loss=0.02567, over 7128.00 frames. ], tot_loss[loss=0.1292, simple_loss=0.2108, pruned_loss=0.02382, over 1441105.49 frames. ], batch size: 98, lr: 3.84e-03, grad_scale: 8.0 +2023-03-21 11:03:01,632 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 11:03:12,017 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 11:03:28,016 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-03-21 11:03:31,239 INFO [train.py:901] (1/2) Epoch 43, batch 1400, loss[loss=0.1288, simple_loss=0.2127, pruned_loss=0.02244, over 7304.00 frames. ], tot_loss[loss=0.1291, simple_loss=0.2105, pruned_loss=0.02386, over 1440905.80 frames. ], batch size: 86, lr: 3.84e-03, grad_scale: 8.0 +2023-03-21 11:03:37,921 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=120021.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:03:46,119 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-21 11:03:49,333 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 11:03:56,946 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.060e+02 1.782e+02 2.120e+02 2.474e+02 4.600e+02, threshold=4.240e+02, percent-clipped=2.0 +2023-03-21 11:03:56,964 INFO [train.py:901] (1/2) Epoch 43, batch 1450, loss[loss=0.1391, simple_loss=0.2257, pruned_loss=0.02623, over 7150.00 frames. ], tot_loss[loss=0.1288, simple_loss=0.2103, pruned_loss=0.02363, over 1440992.54 frames. ], batch size: 98, lr: 3.84e-03, grad_scale: 8.0 +2023-03-21 11:04:14,248 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 11:04:23,464 INFO [train.py:901] (1/2) Epoch 43, batch 1500, loss[loss=0.1469, simple_loss=0.2259, pruned_loss=0.034, over 7127.00 frames. ], tot_loss[loss=0.1288, simple_loss=0.21, pruned_loss=0.02378, over 1443678.93 frames. ], batch size: 98, lr: 3.83e-03, grad_scale: 8.0 +2023-03-21 11:04:26,591 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4590, 4.0887, 4.0972, 4.1449, 4.0956, 4.0298, 4.3976, 3.7603], + device='cuda:1'), covar=tensor([0.0136, 0.0155, 0.0115, 0.0169, 0.0450, 0.0137, 0.0132, 0.0234], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0106, 0.0107, 0.0093, 0.0184, 0.0112, 0.0110, 0.0118], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 11:04:31,031 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 11:04:35,074 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5843, 3.6846, 3.5692, 3.6829, 3.3509, 3.6054, 3.9486, 4.0071], + device='cuda:1'), covar=tensor([0.0230, 0.0180, 0.0255, 0.0179, 0.0416, 0.0409, 0.0255, 0.0184], + device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0127, 0.0121, 0.0126, 0.0113, 0.0102, 0.0098, 0.0101], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 11:04:35,133 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1061, 2.7818, 1.7515, 3.3199, 3.3598, 3.4070, 2.8564, 2.9721], + device='cuda:1'), covar=tensor([0.2356, 0.0967, 0.4724, 0.0495, 0.0264, 0.0236, 0.0352, 0.0314], + device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0230, 0.0241, 0.0257, 0.0199, 0.0203, 0.0216, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:04:43,929 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 +2023-03-21 11:04:48,524 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.608e+02 1.838e+02 2.135e+02 4.788e+02, threshold=3.676e+02, percent-clipped=1.0 +2023-03-21 11:04:48,544 INFO [train.py:901] (1/2) Epoch 43, batch 1550, loss[loss=0.1238, simple_loss=0.2042, pruned_loss=0.02172, over 7333.00 frames. ], tot_loss[loss=0.1285, simple_loss=0.2098, pruned_loss=0.02361, over 1442997.07 frames. ], batch size: 54, lr: 3.83e-03, grad_scale: 8.0 +2023-03-21 11:04:54,082 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 11:04:54,182 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120170.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:05:15,320 INFO [train.py:901] (1/2) Epoch 43, batch 1600, loss[loss=0.1202, simple_loss=0.1994, pruned_loss=0.02052, over 7271.00 frames. ], tot_loss[loss=0.1279, simple_loss=0.2091, pruned_loss=0.02335, over 1443341.90 frames. ], batch size: 52, lr: 3.83e-03, grad_scale: 8.0 +2023-03-21 11:05:19,899 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=120218.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:05:26,389 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 11:05:26,899 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 11:05:30,339 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 11:05:39,546 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 11:05:40,488 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.046e+02 1.635e+02 1.960e+02 2.289e+02 3.677e+02, threshold=3.920e+02, percent-clipped=1.0 +2023-03-21 11:05:40,508 INFO [train.py:901] (1/2) Epoch 43, batch 1650, loss[loss=0.1256, simple_loss=0.2045, pruned_loss=0.02341, over 7259.00 frames. ], tot_loss[loss=0.128, simple_loss=0.2094, pruned_loss=0.02335, over 1441575.20 frames. ], batch size: 47, lr: 3.83e-03, grad_scale: 8.0 +2023-03-21 11:05:44,645 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 11:05:53,370 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 11:06:06,864 INFO [train.py:901] (1/2) Epoch 43, batch 1700, loss[loss=0.127, simple_loss=0.2105, pruned_loss=0.02177, over 7277.00 frames. ], tot_loss[loss=0.128, simple_loss=0.2095, pruned_loss=0.02328, over 1441518.44 frames. ], batch size: 77, lr: 3.83e-03, grad_scale: 8.0 +2023-03-21 11:06:10,363 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 11:06:12,990 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120321.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:06:14,397 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 11:06:22,099 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.3205, 4.7033, 4.6203, 5.3224, 5.0711, 5.1261, 4.7365, 4.7549], + device='cuda:1'), covar=tensor([0.0708, 0.2352, 0.2000, 0.0782, 0.0843, 0.1197, 0.0781, 0.1114], + device='cuda:1'), in_proj_covar=tensor([0.0266, 0.0398, 0.0300, 0.0314, 0.0236, 0.0371, 0.0231, 0.0280], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 11:06:25,046 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 11:06:32,673 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 1.697e+02 1.985e+02 2.246e+02 3.826e+02, threshold=3.970e+02, percent-clipped=0.0 +2023-03-21 11:06:32,691 INFO [train.py:901] (1/2) Epoch 43, batch 1750, loss[loss=0.1535, simple_loss=0.2357, pruned_loss=0.0356, over 7275.00 frames. ], tot_loss[loss=0.1287, simple_loss=0.2103, pruned_loss=0.02357, over 1442111.29 frames. ], batch size: 70, lr: 3.83e-03, grad_scale: 8.0 +2023-03-21 11:06:33,295 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1088, 4.3343, 4.0028, 4.3217, 3.8341, 4.2696, 4.5666, 4.6486], + device='cuda:1'), covar=tensor([0.0237, 0.0146, 0.0238, 0.0170, 0.0394, 0.0283, 0.0220, 0.0167], + device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0129, 0.0123, 0.0128, 0.0115, 0.0103, 0.0100, 0.0103], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 11:06:38,439 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=120369.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:06:49,416 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 11:06:50,416 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 11:06:58,738 INFO [train.py:901] (1/2) Epoch 43, batch 1800, loss[loss=0.1387, simple_loss=0.22, pruned_loss=0.02872, over 7248.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2096, pruned_loss=0.02326, over 1441945.77 frames. ], batch size: 55, lr: 3.83e-03, grad_scale: 8.0 +2023-03-21 11:07:09,908 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9879, 3.6731, 3.6997, 3.6393, 3.6649, 3.4819, 3.8899, 3.4498], + device='cuda:1'), covar=tensor([0.0148, 0.0220, 0.0117, 0.0215, 0.0482, 0.0146, 0.0155, 0.0211], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0106, 0.0107, 0.0092, 0.0184, 0.0112, 0.0110, 0.0118], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 11:07:11,777 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 11:07:24,996 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.732e+02 2.094e+02 2.439e+02 7.280e+02, threshold=4.189e+02, percent-clipped=3.0 +2023-03-21 11:07:25,014 INFO [train.py:901] (1/2) Epoch 43, batch 1850, loss[loss=0.116, simple_loss=0.2031, pruned_loss=0.01444, over 7330.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2099, pruned_loss=0.02321, over 1442656.84 frames. ], batch size: 75, lr: 3.83e-03, grad_scale: 8.0 +2023-03-21 11:07:25,174 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3972, 2.2036, 2.3419, 3.5614, 1.7612, 3.3458, 1.3837, 3.2186], + device='cuda:1'), covar=tensor([0.0221, 0.1627, 0.1943, 0.0300, 0.4404, 0.0347, 0.1366, 0.0376], + device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0243, 0.0254, 0.0211, 0.0248, 0.0218, 0.0220, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:07:25,570 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 11:07:34,608 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3755, 4.1199, 3.6531, 3.8614, 3.2707, 2.3600, 1.7714, 4.2855], + device='cuda:1'), covar=tensor([0.0039, 0.0079, 0.0123, 0.0068, 0.0183, 0.0627, 0.0754, 0.0048], + device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0096, 0.0116, 0.0098, 0.0135, 0.0138, 0.0131, 0.0108], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 11:07:35,983 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 11:07:39,137 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3835, 3.5715, 2.8430, 3.8862, 3.2691, 3.4439, 1.7854, 2.7992], + device='cuda:1'), covar=tensor([0.0461, 0.0855, 0.2652, 0.0470, 0.0595, 0.0603, 0.3961, 0.2065], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0255, 0.0276, 0.0267, 0.0268, 0.0263, 0.0229, 0.0253], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:07:50,003 INFO [train.py:901] (1/2) Epoch 43, batch 1900, loss[loss=0.1448, simple_loss=0.2239, pruned_loss=0.03279, over 7282.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2096, pruned_loss=0.0233, over 1443078.88 frames. ], batch size: 66, lr: 3.83e-03, grad_scale: 8.0 +2023-03-21 11:07:53,033 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 11:08:01,361 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5668, 1.9393, 1.6974, 1.8608, 2.0453, 1.8327, 1.8460, 1.5390], + device='cuda:1'), covar=tensor([0.0132, 0.0173, 0.0245, 0.0158, 0.0095, 0.0135, 0.0121, 0.0197], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0038, 0.0039, 0.0037, 0.0036, 0.0039, 0.0049], + device='cuda:1'), out_proj_covar=tensor([4.5141e-05, 4.4025e-05, 4.2830e-05, 4.3283e-05, 4.1342e-05, 3.9884e-05, + 4.3793e-05, 5.3651e-05], device='cuda:1') +2023-03-21 11:08:03,862 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4737, 2.7557, 2.5879, 2.7525, 2.7593, 2.3195, 2.6355, 2.5328], + device='cuda:1'), covar=tensor([0.0749, 0.0581, 0.0767, 0.0732, 0.0747, 0.0979, 0.0882, 0.0800], + device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0060, 0.0069, 0.0060, 0.0057, 0.0064, 0.0057, 0.0055], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:08:16,188 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.226e+02 1.737e+02 2.080e+02 2.428e+02 3.669e+02, threshold=4.159e+02, percent-clipped=0.0 +2023-03-21 11:08:16,209 INFO [train.py:901] (1/2) Epoch 43, batch 1950, loss[loss=0.09299, simple_loss=0.1545, pruned_loss=0.01573, over 5947.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2098, pruned_loss=0.0233, over 1441844.09 frames. ], batch size: 25, lr: 3.83e-03, grad_scale: 8.0 +2023-03-21 11:08:18,747 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 11:08:30,317 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 11:08:35,382 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 11:08:35,902 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 11:08:41,819 INFO [train.py:901] (1/2) Epoch 43, batch 2000, loss[loss=0.1019, simple_loss=0.1752, pruned_loss=0.01434, over 6979.00 frames. ], tot_loss[loss=0.1276, simple_loss=0.2089, pruned_loss=0.02309, over 1441979.59 frames. ], batch size: 35, lr: 3.83e-03, grad_scale: 8.0 +2023-03-21 11:08:53,534 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 11:08:54,678 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6559, 1.6054, 1.8020, 2.0569, 1.6277, 1.9989, 1.4844, 2.1176], + device='cuda:1'), covar=tensor([0.2035, 0.3292, 0.1512, 0.1631, 0.1096, 0.1227, 0.1847, 0.1449], + device='cuda:1'), in_proj_covar=tensor([0.0081, 0.0081, 0.0075, 0.0068, 0.0067, 0.0066, 0.0106, 0.0069], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:09:05,249 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 11:09:08,204 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.300e+02 1.682e+02 2.036e+02 2.284e+02 3.564e+02, threshold=4.072e+02, percent-clipped=0.0 +2023-03-21 11:09:08,223 INFO [train.py:901] (1/2) Epoch 43, batch 2050, loss[loss=0.1345, simple_loss=0.2208, pruned_loss=0.02409, over 7242.00 frames. ], tot_loss[loss=0.1273, simple_loss=0.2088, pruned_loss=0.02291, over 1442944.44 frames. ], batch size: 55, lr: 3.83e-03, grad_scale: 8.0 +2023-03-21 11:09:13,315 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 11:09:15,424 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3677, 3.9918, 3.9652, 4.0422, 4.0304, 3.8171, 4.2364, 3.7090], + device='cuda:1'), covar=tensor([0.0133, 0.0164, 0.0133, 0.0165, 0.0459, 0.0146, 0.0141, 0.0228], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0106, 0.0106, 0.0092, 0.0183, 0.0112, 0.0109, 0.0118], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 11:09:17,004 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=120676.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:09:32,831 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-03-21 11:09:33,479 INFO [train.py:901] (1/2) Epoch 43, batch 2100, loss[loss=0.1265, simple_loss=0.2097, pruned_loss=0.02166, over 7274.00 frames. ], tot_loss[loss=0.1275, simple_loss=0.2091, pruned_loss=0.02297, over 1443584.74 frames. ], batch size: 70, lr: 3.83e-03, grad_scale: 8.0 +2023-03-21 11:09:46,687 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 11:09:48,884 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=120737.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:09:49,118 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-03-21 11:09:49,726 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 11:09:59,679 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.702e+02 2.039e+02 2.311e+02 3.406e+02, threshold=4.078e+02, percent-clipped=0.0 +2023-03-21 11:09:59,698 INFO [train.py:901] (1/2) Epoch 43, batch 2150, loss[loss=0.1385, simple_loss=0.2226, pruned_loss=0.02726, over 7128.00 frames. ], tot_loss[loss=0.1279, simple_loss=0.2095, pruned_loss=0.02312, over 1442991.06 frames. ], batch size: 98, lr: 3.82e-03, grad_scale: 8.0 +2023-03-21 11:10:26,227 INFO [train.py:901] (1/2) Epoch 43, batch 2200, loss[loss=0.1251, simple_loss=0.2123, pruned_loss=0.019, over 7119.00 frames. ], tot_loss[loss=0.1283, simple_loss=0.2099, pruned_loss=0.0234, over 1441167.56 frames. ], batch size: 98, lr: 3.82e-03, grad_scale: 8.0 +2023-03-21 11:10:32,165 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 +2023-03-21 11:10:34,307 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 11:10:51,388 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.844e+02 2.112e+02 2.583e+02 4.485e+02, threshold=4.224e+02, percent-clipped=2.0 +2023-03-21 11:10:51,407 INFO [train.py:901] (1/2) Epoch 43, batch 2250, loss[loss=0.1564, simple_loss=0.2297, pruned_loss=0.04155, over 7309.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2098, pruned_loss=0.02347, over 1440897.09 frames. ], batch size: 49, lr: 3.82e-03, grad_scale: 8.0 +2023-03-21 11:10:55,775 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 +2023-03-21 11:11:08,262 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 11:11:08,712 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 11:11:17,798 INFO [train.py:901] (1/2) Epoch 43, batch 2300, loss[loss=0.1513, simple_loss=0.2356, pruned_loss=0.03343, over 6757.00 frames. ], tot_loss[loss=0.1283, simple_loss=0.2098, pruned_loss=0.02343, over 1441701.50 frames. ], batch size: 107, lr: 3.82e-03, grad_scale: 8.0 +2023-03-21 11:11:20,310 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 11:11:30,824 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4349, 3.9490, 3.9926, 3.9840, 4.0863, 3.8189, 4.2218, 3.7847], + device='cuda:1'), covar=tensor([0.0136, 0.0192, 0.0135, 0.0208, 0.0423, 0.0144, 0.0153, 0.0203], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0106, 0.0107, 0.0092, 0.0184, 0.0112, 0.0110, 0.0118], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 11:11:40,469 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 +2023-03-21 11:11:42,602 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.126e+02 1.667e+02 2.000e+02 2.396e+02 4.308e+02, threshold=4.000e+02, percent-clipped=1.0 +2023-03-21 11:11:42,620 INFO [train.py:901] (1/2) Epoch 43, batch 2350, loss[loss=0.09619, simple_loss=0.1605, pruned_loss=0.01596, over 6081.00 frames. ], tot_loss[loss=0.1287, simple_loss=0.2103, pruned_loss=0.02356, over 1441999.97 frames. ], batch size: 26, lr: 3.82e-03, grad_scale: 8.0 +2023-03-21 11:12:06,110 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121003.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:12:06,536 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 11:12:09,066 INFO [train.py:901] (1/2) Epoch 43, batch 2400, loss[loss=0.1337, simple_loss=0.2126, pruned_loss=0.02736, over 7245.00 frames. ], tot_loss[loss=0.1297, simple_loss=0.2112, pruned_loss=0.02412, over 1441634.58 frames. ], batch size: 47, lr: 3.82e-03, grad_scale: 8.0 +2023-03-21 11:12:13,068 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 11:12:20,524 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121032.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:12:23,024 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 11:12:25,030 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 11:12:35,172 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.205e+02 1.665e+02 2.063e+02 2.485e+02 4.404e+02, threshold=4.127e+02, percent-clipped=1.0 +2023-03-21 11:12:35,190 INFO [train.py:901] (1/2) Epoch 43, batch 2450, loss[loss=0.1349, simple_loss=0.2228, pruned_loss=0.02353, over 7277.00 frames. ], tot_loss[loss=0.1298, simple_loss=0.2112, pruned_loss=0.0242, over 1442201.30 frames. ], batch size: 57, lr: 3.82e-03, grad_scale: 8.0 +2023-03-21 11:12:37,847 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121064.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:12:51,575 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 +2023-03-21 11:12:51,707 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 11:13:00,333 INFO [train.py:901] (1/2) Epoch 43, batch 2500, loss[loss=0.1229, simple_loss=0.2092, pruned_loss=0.01829, over 7349.00 frames. ], tot_loss[loss=0.1299, simple_loss=0.2114, pruned_loss=0.02415, over 1443797.47 frames. ], batch size: 73, lr: 3.82e-03, grad_scale: 4.0 +2023-03-21 11:13:18,048 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 11:13:26,644 INFO [train.py:901] (1/2) Epoch 43, batch 2550, loss[loss=0.1256, simple_loss=0.2143, pruned_loss=0.01849, over 7260.00 frames. ], tot_loss[loss=0.1297, simple_loss=0.2109, pruned_loss=0.02426, over 1443334.87 frames. ], batch size: 89, lr: 3.82e-03, grad_scale: 4.0 +2023-03-21 11:13:27,090 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.681e+02 1.922e+02 2.296e+02 3.728e+02, threshold=3.843e+02, percent-clipped=0.0 +2023-03-21 11:13:29,395 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121164.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:13:52,441 INFO [train.py:901] (1/2) Epoch 43, batch 2600, loss[loss=0.1232, simple_loss=0.2008, pruned_loss=0.02285, over 7311.00 frames. ], tot_loss[loss=0.1293, simple_loss=0.2101, pruned_loss=0.02424, over 1438666.16 frames. ], batch size: 49, lr: 3.82e-03, grad_scale: 4.0 +2023-03-21 11:14:00,567 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121225.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:14:17,800 INFO [train.py:901] (1/2) Epoch 43, batch 2650, loss[loss=0.1397, simple_loss=0.2178, pruned_loss=0.03078, over 7297.00 frames. ], tot_loss[loss=0.1294, simple_loss=0.21, pruned_loss=0.02438, over 1439222.22 frames. ], batch size: 68, lr: 3.82e-03, grad_scale: 4.0 +2023-03-21 11:14:18,237 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.000e+02 1.753e+02 2.054e+02 2.477e+02 3.622e+02, threshold=4.107e+02, percent-clipped=0.0 +2023-03-21 11:14:32,228 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9197, 2.9320, 2.1430, 3.0907, 2.3406, 2.7860, 1.3507, 2.1411], + device='cuda:1'), covar=tensor([0.0657, 0.1158, 0.3079, 0.0893, 0.0660, 0.0778, 0.4339, 0.1952], + device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0254, 0.0275, 0.0265, 0.0264, 0.0262, 0.0227, 0.0252], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:14:42,686 INFO [train.py:901] (1/2) Epoch 43, batch 2700, loss[loss=0.1096, simple_loss=0.1917, pruned_loss=0.01377, over 7140.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2093, pruned_loss=0.0238, over 1438169.95 frames. ], batch size: 41, lr: 3.82e-03, grad_scale: 4.0 +2023-03-21 11:14:43,842 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121311.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 11:14:54,053 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121332.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:15:06,966 INFO [train.py:901] (1/2) Epoch 43, batch 2750, loss[loss=0.1256, simple_loss=0.2109, pruned_loss=0.02013, over 7242.00 frames. ], tot_loss[loss=0.1287, simple_loss=0.2098, pruned_loss=0.02378, over 1441380.66 frames. ], batch size: 89, lr: 3.81e-03, grad_scale: 4.0 +2023-03-21 11:15:07,044 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121359.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:15:07,425 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.235e+02 1.784e+02 2.196e+02 2.525e+02 4.898e+02, threshold=4.392e+02, percent-clipped=1.0 +2023-03-21 11:15:13,538 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121372.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 11:15:17,308 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=121380.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:15:30,795 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121407.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:15:31,682 INFO [train.py:901] (1/2) Epoch 43, batch 2800, loss[loss=0.09901, simple_loss=0.1735, pruned_loss=0.01227, over 6994.00 frames. ], tot_loss[loss=0.1285, simple_loss=0.2099, pruned_loss=0.02357, over 1443084.16 frames. ], batch size: 35, lr: 3.81e-03, grad_scale: 8.0 +2023-03-21 11:15:35,399 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-03-21 11:15:54,282 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 11:15:54,343 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 11:15:54,370 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 11:15:54,387 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 11:15:54,563 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 11:15:54,579 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 11:15:54,589 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 11:15:54,693 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 11:15:55,091 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 11:15:55,106 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 11:15:55,123 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 11:15:55,290 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 11:15:55,297 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 11:15:55,304 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 11:15:55,446 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 11:15:55,449 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 11:15:55,506 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 11:15:59,601 INFO [train.py:901] (1/2) Epoch 44, batch 0, loss[loss=0.1254, simple_loss=0.2183, pruned_loss=0.01623, over 7353.00 frames. ], tot_loss[loss=0.1254, simple_loss=0.2183, pruned_loss=0.01623, over 7353.00 frames. ], batch size: 73, lr: 3.77e-03, grad_scale: 8.0 +2023-03-21 11:15:59,601 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 11:16:06,949 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3411, 2.5445, 2.9558, 2.4797, 2.4447, 2.4052, 2.3541, 2.0081], + device='cuda:1'), covar=tensor([0.0421, 0.0432, 0.0162, 0.0200, 0.0516, 0.0558, 0.0313, 0.0327], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0040, 0.0041, 0.0040, 0.0037, 0.0037, 0.0044, 0.0043], + device='cuda:1'), out_proj_covar=tensor([1.0406e-04, 1.0404e-04, 1.0254e-04, 1.0218e-04, 9.8559e-05, 9.8310e-05, + 1.0859e-04, 1.0895e-04], device='cuda:1') +2023-03-21 11:16:17,419 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3531, 4.2376, 3.6209, 3.9041, 3.6804, 2.4342, 1.8931, 4.3875], + device='cuda:1'), covar=tensor([0.0044, 0.0069, 0.0122, 0.0079, 0.0125, 0.0673, 0.0738, 0.0045], + device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0094, 0.0115, 0.0097, 0.0133, 0.0136, 0.0129, 0.0106], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 11:16:25,574 INFO [train.py:935] (1/2) Epoch 44, validation: loss=0.1656, simple_loss=0.2575, pruned_loss=0.03679, over 1622729.00 frames. +2023-03-21 11:16:25,575 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 11:16:32,065 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 11:16:39,142 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.245e+02 1.793e+02 2.069e+02 2.336e+02 4.118e+02, threshold=4.138e+02, percent-clipped=0.0 +2023-03-21 11:16:43,105 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 +2023-03-21 11:16:43,287 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 11:16:43,917 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121468.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:16:44,866 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9835, 4.1952, 3.9104, 4.1381, 3.7440, 4.0765, 4.3994, 4.4306], + device='cuda:1'), covar=tensor([0.0248, 0.0154, 0.0268, 0.0197, 0.0386, 0.0285, 0.0248, 0.0188], + device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0130, 0.0123, 0.0129, 0.0117, 0.0104, 0.0101, 0.0105], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 11:16:50,249 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 11:16:51,199 INFO [train.py:901] (1/2) Epoch 44, batch 50, loss[loss=0.1339, simple_loss=0.2185, pruned_loss=0.02462, over 7204.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.2089, pruned_loss=0.02323, over 327371.98 frames. ], batch size: 93, lr: 3.77e-03, grad_scale: 8.0 +2023-03-21 11:16:52,214 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 11:16:55,160 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 11:17:08,018 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6390, 5.1727, 5.2488, 5.1169, 4.9912, 4.6609, 5.2985, 5.0471], + device='cuda:1'), covar=tensor([0.0456, 0.0363, 0.0409, 0.0586, 0.0374, 0.0419, 0.0297, 0.0497], + device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0265, 0.0205, 0.0206, 0.0159, 0.0231, 0.0214, 0.0151], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:17:10,457 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121520.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:17:12,875 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 11:17:13,368 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 11:17:16,910 INFO [train.py:901] (1/2) Epoch 44, batch 100, loss[loss=0.1245, simple_loss=0.2098, pruned_loss=0.01962, over 7377.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2078, pruned_loss=0.02254, over 576153.24 frames. ], batch size: 65, lr: 3.77e-03, grad_scale: 8.0 +2023-03-21 11:17:17,104 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1814, 3.1734, 2.9892, 2.9923, 2.6689, 2.7431, 3.2464, 2.1785], + device='cuda:1'), covar=tensor([0.0853, 0.0846, 0.0790, 0.0952, 0.1089, 0.1522, 0.0855, 0.3196], + device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0335, 0.0270, 0.0351, 0.0282, 0.0286, 0.0346, 0.0239], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:17:20,783 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-03-21 11:17:29,614 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7538, 2.4724, 2.5409, 3.6992, 2.0322, 3.4079, 1.4038, 3.2865], + device='cuda:1'), covar=tensor([0.0212, 0.1556, 0.1865, 0.0216, 0.4126, 0.0284, 0.1422, 0.0529], + device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0243, 0.0254, 0.0208, 0.0248, 0.0216, 0.0220, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:17:30,939 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.151e+02 1.757e+02 2.045e+02 2.523e+02 5.240e+02, threshold=4.089e+02, percent-clipped=2.0 +2023-03-21 11:17:42,367 INFO [train.py:901] (1/2) Epoch 44, batch 150, loss[loss=0.1295, simple_loss=0.2119, pruned_loss=0.0235, over 7321.00 frames. ], tot_loss[loss=0.126, simple_loss=0.2078, pruned_loss=0.02212, over 769134.58 frames. ], batch size: 54, lr: 3.77e-03, grad_scale: 8.0 +2023-03-21 11:18:08,080 INFO [train.py:901] (1/2) Epoch 44, batch 200, loss[loss=0.1246, simple_loss=0.1992, pruned_loss=0.025, over 7287.00 frames. ], tot_loss[loss=0.1263, simple_loss=0.2077, pruned_loss=0.02245, over 920702.86 frames. ], batch size: 66, lr: 3.77e-03, grad_scale: 8.0 +2023-03-21 11:18:13,027 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-03-21 11:18:13,172 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 11:18:18,141 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 11:18:21,724 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121659.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:18:22,112 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.731e+02 1.966e+02 2.268e+02 4.894e+02, threshold=3.933e+02, percent-clipped=1.0 +2023-03-21 11:18:23,128 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 11:18:25,633 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121667.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 11:18:34,220 INFO [train.py:901] (1/2) Epoch 44, batch 250, loss[loss=0.1121, simple_loss=0.1931, pruned_loss=0.01552, over 7330.00 frames. ], tot_loss[loss=0.1274, simple_loss=0.2091, pruned_loss=0.02287, over 1037443.07 frames. ], batch size: 75, lr: 3.77e-03, grad_scale: 8.0 +2023-03-21 11:18:36,706 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 11:18:43,125 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-21 11:18:46,463 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=121707.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:18:50,886 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 +2023-03-21 11:18:57,507 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 11:18:59,994 INFO [train.py:901] (1/2) Epoch 44, batch 300, loss[loss=0.1088, simple_loss=0.1915, pruned_loss=0.01307, over 7345.00 frames. ], tot_loss[loss=0.127, simple_loss=0.2085, pruned_loss=0.02276, over 1123975.44 frames. ], batch size: 44, lr: 3.77e-03, grad_scale: 8.0 +2023-03-21 11:19:04,979 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 11:19:13,966 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.695e+02 1.999e+02 2.346e+02 5.920e+02, threshold=3.997e+02, percent-clipped=4.0 +2023-03-21 11:19:15,572 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121763.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:19:25,447 INFO [train.py:901] (1/2) Epoch 44, batch 350, loss[loss=0.1495, simple_loss=0.2309, pruned_loss=0.03408, over 7269.00 frames. ], tot_loss[loss=0.1279, simple_loss=0.2096, pruned_loss=0.02308, over 1195045.26 frames. ], batch size: 57, lr: 3.76e-03, grad_scale: 8.0 +2023-03-21 11:19:34,467 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5492, 4.1742, 4.1450, 4.2280, 4.2480, 4.1172, 4.4896, 3.8851], + device='cuda:1'), covar=tensor([0.0128, 0.0149, 0.0118, 0.0141, 0.0372, 0.0099, 0.0107, 0.0179], + device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0107, 0.0108, 0.0094, 0.0187, 0.0113, 0.0110, 0.0120], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 11:19:38,792 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-03-21 11:19:40,897 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 11:19:41,745 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 +2023-03-21 11:19:44,968 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121820.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:19:51,297 INFO [train.py:901] (1/2) Epoch 44, batch 400, loss[loss=0.1233, simple_loss=0.2031, pruned_loss=0.02176, over 7351.00 frames. ], tot_loss[loss=0.1285, simple_loss=0.2103, pruned_loss=0.02334, over 1251363.13 frames. ], batch size: 73, lr: 3.76e-03, grad_scale: 8.0 +2023-03-21 11:20:05,403 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.230e+02 1.734e+02 2.013e+02 2.477e+02 7.219e+02, threshold=4.025e+02, percent-clipped=2.0 +2023-03-21 11:20:09,536 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=121868.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:20:17,012 INFO [train.py:901] (1/2) Epoch 44, batch 450, loss[loss=0.1201, simple_loss=0.1971, pruned_loss=0.02156, over 7313.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2098, pruned_loss=0.02325, over 1292814.30 frames. ], batch size: 49, lr: 3.76e-03, grad_scale: 8.0 +2023-03-21 11:20:22,015 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 11:20:22,476 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 11:20:31,200 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1605, 4.1354, 3.3936, 3.6879, 3.0813, 2.2225, 1.7958, 4.1830], + device='cuda:1'), covar=tensor([0.0053, 0.0062, 0.0138, 0.0094, 0.0203, 0.0672, 0.0734, 0.0050], + device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0094, 0.0115, 0.0098, 0.0134, 0.0136, 0.0129, 0.0107], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 11:20:43,171 INFO [train.py:901] (1/2) Epoch 44, batch 500, loss[loss=0.1089, simple_loss=0.1819, pruned_loss=0.01796, over 7002.00 frames. ], tot_loss[loss=0.128, simple_loss=0.2097, pruned_loss=0.0232, over 1326846.52 frames. ], batch size: 35, lr: 3.76e-03, grad_scale: 8.0 +2023-03-21 11:20:51,971 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121950.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:20:54,404 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 11:20:55,817 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 11:20:56,388 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 11:20:56,847 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.208e+02 1.686e+02 1.981e+02 2.363e+02 5.254e+02, threshold=3.961e+02, percent-clipped=1.0 +2023-03-21 11:20:58,417 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 11:21:00,586 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121967.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 11:21:02,934 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 11:21:06,089 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7970, 2.9527, 3.7390, 3.7637, 3.7766, 3.7929, 3.7029, 3.6925], + device='cuda:1'), covar=tensor([0.0030, 0.0138, 0.0036, 0.0037, 0.0034, 0.0034, 0.0054, 0.0047], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0072, 0.0060, 0.0058, 0.0056, 0.0061, 0.0049, 0.0080], + device='cuda:1'), out_proj_covar=tensor([8.2712e-05, 1.4391e-04, 1.0561e-04, 9.7637e-05, 9.3228e-05, 1.0553e-04, + 9.1815e-05, 1.4575e-04], device='cuda:1') +2023-03-21 11:21:09,135 INFO [train.py:901] (1/2) Epoch 44, batch 550, loss[loss=0.1294, simple_loss=0.214, pruned_loss=0.02236, over 7367.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2102, pruned_loss=0.0233, over 1354503.73 frames. ], batch size: 63, lr: 3.76e-03, grad_scale: 8.0 +2023-03-21 11:21:15,029 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-03-21 11:21:15,191 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 11:21:23,699 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 11:21:24,438 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122011.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:21:26,330 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=122015.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 11:21:26,431 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3694, 4.1218, 3.9362, 3.9824, 3.5371, 3.4505, 4.3650, 2.7952], + device='cuda:1'), covar=tensor([0.0553, 0.0816, 0.0561, 0.0888, 0.1182, 0.1441, 0.0812, 0.3155], + device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0337, 0.0271, 0.0352, 0.0282, 0.0287, 0.0348, 0.0240], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:21:27,743 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 11:21:34,155 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 11:21:35,132 INFO [train.py:901] (1/2) Epoch 44, batch 600, loss[loss=0.1151, simple_loss=0.2005, pruned_loss=0.01486, over 7209.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2099, pruned_loss=0.02313, over 1374128.02 frames. ], batch size: 50, lr: 3.76e-03, grad_scale: 8.0 +2023-03-21 11:21:39,853 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0492, 2.5849, 2.6971, 4.0249, 2.0233, 3.8517, 1.5382, 3.6397], + device='cuda:1'), covar=tensor([0.0180, 0.1464, 0.1843, 0.0203, 0.4513, 0.0255, 0.1356, 0.0381], + device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0240, 0.0251, 0.0205, 0.0244, 0.0213, 0.0218, 0.0222], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:21:48,864 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 1.747e+02 1.962e+02 2.510e+02 3.500e+02, threshold=3.924e+02, percent-clipped=0.0 +2023-03-21 11:21:50,416 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 11:21:50,500 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122063.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:22:00,077 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 11:22:01,116 INFO [train.py:901] (1/2) Epoch 44, batch 650, loss[loss=0.1308, simple_loss=0.2193, pruned_loss=0.02112, over 7314.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.21, pruned_loss=0.02342, over 1387724.57 frames. ], batch size: 49, lr: 3.76e-03, grad_scale: 8.0 +2023-03-21 11:22:15,968 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=122111.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:22:18,045 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 11:22:19,133 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122117.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:22:26,059 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 11:22:27,019 INFO [train.py:901] (1/2) Epoch 44, batch 700, loss[loss=0.1222, simple_loss=0.1987, pruned_loss=0.02287, over 7352.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2094, pruned_loss=0.02371, over 1397079.57 frames. ], batch size: 63, lr: 3.76e-03, grad_scale: 8.0 +2023-03-21 11:22:41,311 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.249e+02 1.853e+02 2.110e+02 2.490e+02 4.389e+02, threshold=4.220e+02, percent-clipped=2.0 +2023-03-21 11:22:50,411 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 11:22:50,550 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122178.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 11:22:50,914 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 11:22:52,893 INFO [train.py:901] (1/2) Epoch 44, batch 750, loss[loss=0.1266, simple_loss=0.2105, pruned_loss=0.02134, over 7289.00 frames. ], tot_loss[loss=0.1286, simple_loss=0.2099, pruned_loss=0.02369, over 1405364.42 frames. ], batch size: 70, lr: 3.76e-03, grad_scale: 8.0 +2023-03-21 11:22:56,516 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7189, 2.9186, 2.1777, 3.1459, 2.3154, 2.5816, 1.3878, 2.2740], + device='cuda:1'), covar=tensor([0.0491, 0.0847, 0.3129, 0.0846, 0.0473, 0.0672, 0.4393, 0.1983], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0256, 0.0277, 0.0267, 0.0268, 0.0264, 0.0229, 0.0254], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:22:59,474 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122195.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:23:00,467 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8763, 1.4503, 2.1292, 2.4447, 2.1099, 2.1535, 2.0699, 2.4280], + device='cuda:1'), covar=tensor([0.2221, 0.2864, 0.2565, 0.1736, 0.2368, 0.3825, 0.1708, 0.2183], + device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0083, 0.0077, 0.0069, 0.0068, 0.0068, 0.0109, 0.0071], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:23:05,681 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 11:23:10,192 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 11:23:16,221 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 11:23:17,238 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 11:23:18,707 INFO [train.py:901] (1/2) Epoch 44, batch 800, loss[loss=0.1333, simple_loss=0.2164, pruned_loss=0.02505, over 7247.00 frames. ], tot_loss[loss=0.1288, simple_loss=0.2105, pruned_loss=0.02354, over 1414881.39 frames. ], batch size: 64, lr: 3.76e-03, grad_scale: 8.0 +2023-03-21 11:23:28,384 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 11:23:29,603 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3585, 2.8784, 2.1599, 3.1945, 3.1940, 3.0121, 2.9094, 2.7017], + device='cuda:1'), covar=tensor([0.2066, 0.1052, 0.3675, 0.0598, 0.0308, 0.0290, 0.0361, 0.0404], + device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0232, 0.0245, 0.0260, 0.0202, 0.0207, 0.0219, 0.0231], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:23:31,227 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122256.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:23:33,116 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.265e+02 1.730e+02 1.943e+02 2.255e+02 3.893e+02, threshold=3.886e+02, percent-clipped=0.0 +2023-03-21 11:23:45,354 INFO [train.py:901] (1/2) Epoch 44, batch 850, loss[loss=0.1019, simple_loss=0.1673, pruned_loss=0.01823, over 6060.00 frames. ], tot_loss[loss=0.1285, simple_loss=0.21, pruned_loss=0.02348, over 1418055.00 frames. ], batch size: 26, lr: 3.76e-03, grad_scale: 8.0 +2023-03-21 11:23:48,858 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 11:23:48,868 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 11:23:54,399 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 11:23:56,989 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122306.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:23:57,936 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 11:24:03,552 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122319.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:24:10,462 INFO [train.py:901] (1/2) Epoch 44, batch 900, loss[loss=0.1151, simple_loss=0.1954, pruned_loss=0.01736, over 7139.00 frames. ], tot_loss[loss=0.1276, simple_loss=0.209, pruned_loss=0.02307, over 1423850.74 frames. ], batch size: 39, lr: 3.76e-03, grad_scale: 8.0 +2023-03-21 11:24:24,638 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.690e+02 1.979e+02 2.336e+02 1.057e+03, threshold=3.958e+02, percent-clipped=2.0 +2023-03-21 11:24:24,839 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2801, 3.5218, 2.5233, 3.8489, 3.0783, 3.2625, 1.6381, 2.5969], + device='cuda:1'), covar=tensor([0.0419, 0.1066, 0.2683, 0.0489, 0.0552, 0.0577, 0.3964, 0.1959], + device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0255, 0.0276, 0.0266, 0.0266, 0.0263, 0.0227, 0.0254], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:24:35,451 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122380.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:24:36,815 INFO [train.py:901] (1/2) Epoch 44, batch 950, loss[loss=0.1276, simple_loss=0.2074, pruned_loss=0.0239, over 7200.00 frames. ], tot_loss[loss=0.1276, simple_loss=0.2091, pruned_loss=0.02307, over 1429816.70 frames. ], batch size: 50, lr: 3.76e-03, grad_scale: 8.0 +2023-03-21 11:24:37,868 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 11:25:01,739 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 11:25:02,668 INFO [train.py:901] (1/2) Epoch 44, batch 1000, loss[loss=0.1278, simple_loss=0.2149, pruned_loss=0.02031, over 7329.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2094, pruned_loss=0.02336, over 1432091.74 frames. ], batch size: 61, lr: 3.75e-03, grad_scale: 8.0 +2023-03-21 11:25:16,931 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.044e+02 1.750e+02 2.006e+02 2.317e+02 3.184e+02, threshold=4.012e+02, percent-clipped=0.0 +2023-03-21 11:25:21,130 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122468.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:25:23,518 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 11:25:23,591 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122473.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 11:25:28,431 INFO [train.py:901] (1/2) Epoch 44, batch 1050, loss[loss=0.1294, simple_loss=0.2147, pruned_loss=0.0221, over 7283.00 frames. ], tot_loss[loss=0.1275, simple_loss=0.2087, pruned_loss=0.02321, over 1432083.61 frames. ], batch size: 68, lr: 3.75e-03, grad_scale: 8.0 +2023-03-21 11:25:45,268 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 11:25:50,598 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 11:25:52,756 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122529.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:25:54,634 INFO [train.py:901] (1/2) Epoch 44, batch 1100, loss[loss=0.1355, simple_loss=0.2178, pruned_loss=0.02656, over 7246.00 frames. ], tot_loss[loss=0.1279, simple_loss=0.2092, pruned_loss=0.02335, over 1436659.01 frames. ], batch size: 55, lr: 3.75e-03, grad_scale: 8.0 +2023-03-21 11:26:02,112 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.49 vs. limit=5.0 +2023-03-21 11:26:04,400 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122551.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:26:08,863 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.256e+02 1.775e+02 1.996e+02 2.337e+02 4.122e+02, threshold=3.992e+02, percent-clipped=0.0 +2023-03-21 11:26:11,010 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9361, 3.8797, 3.1411, 3.5221, 2.9897, 2.0928, 1.7976, 3.9239], + device='cuda:1'), covar=tensor([0.0080, 0.0078, 0.0213, 0.0108, 0.0250, 0.0805, 0.0810, 0.0088], + device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0094, 0.0115, 0.0098, 0.0132, 0.0135, 0.0128, 0.0106], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 11:26:18,407 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 11:26:18,948 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 11:26:20,442 INFO [train.py:901] (1/2) Epoch 44, batch 1150, loss[loss=0.1243, simple_loss=0.2048, pruned_loss=0.0219, over 7147.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2097, pruned_loss=0.02329, over 1438145.36 frames. ], batch size: 41, lr: 3.75e-03, grad_scale: 8.0 +2023-03-21 11:26:24,684 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2064, 3.1402, 3.2745, 3.2996, 2.9499, 2.8917, 3.3636, 2.4089], + device='cuda:1'), covar=tensor([0.0523, 0.0689, 0.0851, 0.0690, 0.0784, 0.1102, 0.0694, 0.2692], + device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0337, 0.0271, 0.0353, 0.0283, 0.0286, 0.0347, 0.0239], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:26:32,124 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3218, 4.7584, 4.8468, 4.7616, 4.6895, 4.3182, 4.8645, 4.6815], + device='cuda:1'), covar=tensor([0.0457, 0.0404, 0.0358, 0.0517, 0.0336, 0.0433, 0.0309, 0.0445], + device='cuda:1'), in_proj_covar=tensor([0.0220, 0.0267, 0.0205, 0.0205, 0.0159, 0.0232, 0.0214, 0.0151], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:26:32,553 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 11:26:33,135 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122606.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:26:33,523 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 11:26:41,051 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 +2023-03-21 11:26:47,171 INFO [train.py:901] (1/2) Epoch 44, batch 1200, loss[loss=0.1274, simple_loss=0.2086, pruned_loss=0.02309, over 7309.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2094, pruned_loss=0.02342, over 1437955.41 frames. ], batch size: 80, lr: 3.75e-03, grad_scale: 8.0 +2023-03-21 11:26:57,732 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=122654.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:26:59,360 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9544, 1.5502, 2.1538, 2.3577, 2.1795, 2.3972, 2.1131, 2.4321], + device='cuda:1'), covar=tensor([0.2581, 0.5378, 0.1637, 0.2088, 0.1112, 0.3227, 0.2330, 0.2280], + device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0082, 0.0075, 0.0068, 0.0067, 0.0066, 0.0107, 0.0070], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:27:00,710 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 1.642e+02 1.822e+02 2.159e+02 3.974e+02, threshold=3.644e+02, percent-clipped=1.0 +2023-03-21 11:27:05,246 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 11:27:08,246 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122675.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:27:12,258 INFO [train.py:901] (1/2) Epoch 44, batch 1250, loss[loss=0.1276, simple_loss=0.2093, pruned_loss=0.02295, over 7329.00 frames. ], tot_loss[loss=0.1289, simple_loss=0.2103, pruned_loss=0.02374, over 1439832.29 frames. ], batch size: 75, lr: 3.75e-03, grad_scale: 8.0 +2023-03-21 11:27:21,513 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1171, 1.7039, 2.2014, 2.4986, 2.4009, 2.4800, 2.3942, 2.4787], + device='cuda:1'), covar=tensor([0.2798, 0.3674, 0.2120, 0.2095, 0.1891, 0.2188, 0.1804, 0.3728], + device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0082, 0.0075, 0.0068, 0.0067, 0.0066, 0.0107, 0.0070], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:27:28,452 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 11:27:33,688 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 11:27:34,748 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 11:27:38,714 INFO [train.py:901] (1/2) Epoch 44, batch 1300, loss[loss=0.1306, simple_loss=0.2115, pruned_loss=0.02485, over 7270.00 frames. ], tot_loss[loss=0.1286, simple_loss=0.21, pruned_loss=0.02359, over 1440689.77 frames. ], batch size: 77, lr: 3.75e-03, grad_scale: 8.0 +2023-03-21 11:27:52,210 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.08 vs. limit=5.0 +2023-03-21 11:27:52,321 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.777e+02 1.977e+02 2.285e+02 3.721e+02, threshold=3.954e+02, percent-clipped=1.0 +2023-03-21 11:27:57,953 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 11:27:59,008 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122773.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 11:28:00,495 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 11:28:00,751 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 +2023-03-21 11:28:03,480 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 11:28:04,458 INFO [train.py:901] (1/2) Epoch 44, batch 1350, loss[loss=0.1098, simple_loss=0.1908, pruned_loss=0.01438, over 7169.00 frames. ], tot_loss[loss=0.1283, simple_loss=0.21, pruned_loss=0.02327, over 1442643.80 frames. ], batch size: 39, lr: 3.75e-03, grad_scale: 8.0 +2023-03-21 11:28:04,600 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5559, 4.3795, 4.0575, 4.0350, 3.8081, 2.6798, 2.2694, 4.6110], + device='cuda:1'), covar=tensor([0.0045, 0.0056, 0.0095, 0.0068, 0.0117, 0.0514, 0.0557, 0.0046], + device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0093, 0.0115, 0.0097, 0.0132, 0.0135, 0.0128, 0.0106], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 11:28:07,163 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0362, 2.2760, 1.8282, 2.8257, 2.8683, 2.7970, 2.7190, 2.6386], + device='cuda:1'), covar=tensor([0.2402, 0.1358, 0.4131, 0.1222, 0.0534, 0.0499, 0.0608, 0.0618], + device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0234, 0.0246, 0.0261, 0.0203, 0.0207, 0.0220, 0.0233], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:28:13,686 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 11:28:18,396 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-21 11:28:24,620 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=122821.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:28:26,163 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122824.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:28:30,641 INFO [train.py:901] (1/2) Epoch 44, batch 1400, loss[loss=0.1381, simple_loss=0.2189, pruned_loss=0.02867, over 7298.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2098, pruned_loss=0.02318, over 1442292.19 frames. ], batch size: 77, lr: 3.75e-03, grad_scale: 8.0 +2023-03-21 11:28:39,851 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122851.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:28:44,868 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.199e+02 1.709e+02 2.025e+02 2.415e+02 3.679e+02, threshold=4.051e+02, percent-clipped=0.0 +2023-03-21 11:28:46,386 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 11:28:51,062 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122872.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:28:56,501 INFO [train.py:901] (1/2) Epoch 44, batch 1450, loss[loss=0.1074, simple_loss=0.1784, pruned_loss=0.01813, over 7163.00 frames. ], tot_loss[loss=0.1274, simple_loss=0.2091, pruned_loss=0.02284, over 1443882.83 frames. ], batch size: 39, lr: 3.75e-03, grad_scale: 8.0 +2023-03-21 11:29:05,324 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=122899.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:29:10,780 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 11:29:18,623 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-03-21 11:29:22,309 INFO [train.py:901] (1/2) Epoch 44, batch 1500, loss[loss=0.1375, simple_loss=0.2227, pruned_loss=0.02622, over 7256.00 frames. ], tot_loss[loss=0.1279, simple_loss=0.2097, pruned_loss=0.02302, over 1444908.18 frames. ], batch size: 55, lr: 3.75e-03, grad_scale: 8.0 +2023-03-21 11:29:22,450 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122933.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:29:26,354 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 11:29:35,926 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 +2023-03-21 11:29:36,567 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.737e+02 1.988e+02 2.376e+02 3.622e+02, threshold=3.976e+02, percent-clipped=0.0 +2023-03-21 11:29:44,774 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122975.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:29:47,858 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1282, 2.4851, 1.9013, 2.7018, 2.8885, 2.9303, 2.5331, 2.4487], + device='cuda:1'), covar=tensor([0.2413, 0.1262, 0.4116, 0.0899, 0.0498, 0.0499, 0.0481, 0.0499], + device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0231, 0.0243, 0.0257, 0.0202, 0.0206, 0.0218, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:29:48,674 INFO [train.py:901] (1/2) Epoch 44, batch 1550, loss[loss=0.1219, simple_loss=0.2079, pruned_loss=0.01792, over 7263.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2099, pruned_loss=0.02322, over 1445987.48 frames. ], batch size: 64, lr: 3.75e-03, grad_scale: 8.0 +2023-03-21 11:29:50,220 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 11:30:09,330 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=123023.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:30:14,954 INFO [train.py:901] (1/2) Epoch 44, batch 1600, loss[loss=0.1244, simple_loss=0.2028, pruned_loss=0.02301, over 7281.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2097, pruned_loss=0.02327, over 1444581.93 frames. ], batch size: 57, lr: 3.75e-03, grad_scale: 8.0 +2023-03-21 11:30:22,713 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 11:30:23,696 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 11:30:26,721 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 11:30:29,318 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.296e+02 1.717e+02 1.945e+02 2.316e+02 3.893e+02, threshold=3.890e+02, percent-clipped=0.0 +2023-03-21 11:30:36,907 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 11:30:39,712 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 +2023-03-21 11:30:41,000 INFO [train.py:901] (1/2) Epoch 44, batch 1650, loss[loss=0.1183, simple_loss=0.1978, pruned_loss=0.01937, over 7144.00 frames. ], tot_loss[loss=0.1279, simple_loss=0.2092, pruned_loss=0.02327, over 1442431.99 frames. ], batch size: 41, lr: 3.74e-03, grad_scale: 16.0 +2023-03-21 11:30:41,492 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 11:30:50,109 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 11:30:57,595 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 +2023-03-21 11:31:02,460 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123124.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:31:05,465 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1669, 3.1151, 2.2288, 3.5858, 2.6355, 2.8052, 1.4588, 2.4436], + device='cuda:1'), covar=tensor([0.0556, 0.1009, 0.3035, 0.0679, 0.0622, 0.0560, 0.4522, 0.1903], + device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0256, 0.0276, 0.0268, 0.0267, 0.0264, 0.0228, 0.0255], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:31:06,784 INFO [train.py:901] (1/2) Epoch 44, batch 1700, loss[loss=0.1506, simple_loss=0.2278, pruned_loss=0.03671, over 7359.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2096, pruned_loss=0.02362, over 1443597.17 frames. ], batch size: 54, lr: 3.74e-03, grad_scale: 16.0 +2023-03-21 11:31:07,731 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 11:31:12,298 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 11:31:21,065 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.227e+02 1.700e+02 1.935e+02 2.300e+02 4.091e+02, threshold=3.869e+02, percent-clipped=2.0 +2023-03-21 11:31:23,520 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 11:31:27,136 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=123172.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:31:32,593 INFO [train.py:901] (1/2) Epoch 44, batch 1750, loss[loss=0.1253, simple_loss=0.2129, pruned_loss=0.01885, over 7305.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2098, pruned_loss=0.02354, over 1446229.44 frames. ], batch size: 80, lr: 3.74e-03, grad_scale: 16.0 +2023-03-21 11:31:48,855 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 11:31:49,891 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 11:31:51,466 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.8439, 5.2922, 5.3884, 5.3409, 5.0253, 4.8452, 5.4220, 5.2142], + device='cuda:1'), covar=tensor([0.0405, 0.0349, 0.0338, 0.0403, 0.0356, 0.0345, 0.0269, 0.0378], + device='cuda:1'), in_proj_covar=tensor([0.0220, 0.0268, 0.0205, 0.0204, 0.0159, 0.0233, 0.0214, 0.0151], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:31:56,434 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123228.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:31:58,859 INFO [train.py:901] (1/2) Epoch 44, batch 1800, loss[loss=0.1357, simple_loss=0.2168, pruned_loss=0.02731, over 7327.00 frames. ], tot_loss[loss=0.1278, simple_loss=0.2093, pruned_loss=0.02309, over 1446619.48 frames. ], batch size: 75, lr: 3.74e-03, grad_scale: 8.0 +2023-03-21 11:32:11,096 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 11:32:13,579 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.238e+02 1.759e+02 1.926e+02 2.309e+02 3.312e+02, threshold=3.851e+02, percent-clipped=0.0 +2023-03-21 11:32:24,076 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 11:32:24,492 INFO [train.py:901] (1/2) Epoch 44, batch 1850, loss[loss=0.1462, simple_loss=0.2238, pruned_loss=0.03433, over 7260.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2098, pruned_loss=0.0233, over 1444819.27 frames. ], batch size: 64, lr: 3.74e-03, grad_scale: 8.0 +2023-03-21 11:32:34,157 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 +2023-03-21 11:32:35,355 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 11:32:40,985 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123314.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:32:51,195 INFO [train.py:901] (1/2) Epoch 44, batch 1900, loss[loss=0.1326, simple_loss=0.2162, pruned_loss=0.02453, over 7353.00 frames. ], tot_loss[loss=0.128, simple_loss=0.2094, pruned_loss=0.02326, over 1444925.92 frames. ], batch size: 73, lr: 3.74e-03, grad_scale: 8.0 +2023-03-21 11:32:52,747 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 11:32:55,060 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 +2023-03-21 11:33:05,244 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.291e+02 1.842e+02 2.137e+02 2.447e+02 5.205e+02, threshold=4.273e+02, percent-clipped=1.0 +2023-03-21 11:33:12,490 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123375.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:33:15,718 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 11:33:16,211 INFO [train.py:901] (1/2) Epoch 44, batch 1950, loss[loss=0.1448, simple_loss=0.2305, pruned_loss=0.02956, over 6803.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2098, pruned_loss=0.02321, over 1443843.94 frames. ], batch size: 107, lr: 3.74e-03, grad_scale: 8.0 +2023-03-21 11:33:26,301 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9369, 3.0665, 3.9189, 3.8167, 3.9248, 3.9370, 3.9570, 3.7857], + device='cuda:1'), covar=tensor([0.0027, 0.0135, 0.0033, 0.0030, 0.0029, 0.0026, 0.0038, 0.0049], + device='cuda:1'), in_proj_covar=tensor([0.0050, 0.0072, 0.0059, 0.0058, 0.0055, 0.0060, 0.0049, 0.0080], + device='cuda:1'), out_proj_covar=tensor([8.2529e-05, 1.4356e-04, 1.0506e-04, 9.7570e-05, 9.2349e-05, 1.0259e-04, + 9.1372e-05, 1.4455e-04], device='cuda:1') +2023-03-21 11:33:27,723 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 11:33:33,364 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 11:33:34,321 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 11:33:40,522 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123428.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:33:42,916 INFO [train.py:901] (1/2) Epoch 44, batch 2000, loss[loss=0.1369, simple_loss=0.2257, pruned_loss=0.0241, over 7159.00 frames. ], tot_loss[loss=0.1286, simple_loss=0.21, pruned_loss=0.02357, over 1445173.59 frames. ], batch size: 98, lr: 3.74e-03, grad_scale: 8.0 +2023-03-21 11:33:46,113 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2125, 2.4441, 2.6526, 2.2366, 2.5110, 2.1742, 2.2165, 1.8861], + device='cuda:1'), covar=tensor([0.0577, 0.0533, 0.0318, 0.0436, 0.0611, 0.0529, 0.0371, 0.0408], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0041, 0.0041, 0.0041, 0.0038, 0.0038, 0.0044, 0.0044], + device='cuda:1'), out_proj_covar=tensor([1.0538e-04, 1.0518e-04, 1.0486e-04, 1.0456e-04, 1.0072e-04, 9.9542e-05, + 1.0996e-04, 1.1063e-04], device='cuda:1') +2023-03-21 11:33:47,761 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.44 vs. limit=5.0 +2023-03-21 11:33:50,452 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 11:33:56,967 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.089e+02 1.719e+02 1.918e+02 2.343e+02 3.686e+02, threshold=3.836e+02, percent-clipped=0.0 +2023-03-21 11:34:01,599 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 11:34:08,696 INFO [train.py:901] (1/2) Epoch 44, batch 2050, loss[loss=0.127, simple_loss=0.2104, pruned_loss=0.02186, over 7288.00 frames. ], tot_loss[loss=0.1283, simple_loss=0.2097, pruned_loss=0.02349, over 1444914.09 frames. ], batch size: 70, lr: 3.74e-03, grad_scale: 8.0 +2023-03-21 11:34:10,725 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 11:34:11,880 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123489.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:34:13,518 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0966, 2.4606, 1.9427, 2.8555, 2.8717, 3.0105, 2.5430, 2.4988], + device='cuda:1'), covar=tensor([0.2321, 0.1206, 0.3784, 0.0937, 0.0456, 0.0423, 0.0491, 0.0462], + device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0232, 0.0244, 0.0260, 0.0202, 0.0207, 0.0220, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:34:32,080 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123528.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:34:34,496 INFO [train.py:901] (1/2) Epoch 44, batch 2100, loss[loss=0.1281, simple_loss=0.2101, pruned_loss=0.02299, over 7294.00 frames. ], tot_loss[loss=0.1289, simple_loss=0.2105, pruned_loss=0.02365, over 1444448.88 frames. ], batch size: 57, lr: 3.74e-03, grad_scale: 8.0 +2023-03-21 11:34:40,136 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3989, 4.2400, 3.4505, 3.8730, 3.2746, 2.2864, 1.7962, 4.3827], + device='cuda:1'), covar=tensor([0.0039, 0.0053, 0.0128, 0.0069, 0.0156, 0.0608, 0.0652, 0.0040], + device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0094, 0.0116, 0.0099, 0.0133, 0.0137, 0.0130, 0.0107], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 11:34:40,207 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4962, 3.2572, 3.4147, 3.4003, 3.1472, 2.9966, 3.6034, 2.5554], + device='cuda:1'), covar=tensor([0.0544, 0.0658, 0.0709, 0.0694, 0.0646, 0.1028, 0.0688, 0.2554], + device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0340, 0.0274, 0.0355, 0.0285, 0.0289, 0.0351, 0.0242], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:34:42,626 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 11:34:45,128 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 11:34:48,573 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 1.796e+02 2.189e+02 2.494e+02 4.694e+02, threshold=4.378e+02, percent-clipped=1.0 +2023-03-21 11:34:56,964 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=123576.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:35:00,428 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 +2023-03-21 11:35:01,053 INFO [train.py:901] (1/2) Epoch 44, batch 2150, loss[loss=0.1132, simple_loss=0.1996, pruned_loss=0.01343, over 7244.00 frames. ], tot_loss[loss=0.1287, simple_loss=0.2102, pruned_loss=0.02357, over 1442143.34 frames. ], batch size: 55, lr: 3.74e-03, grad_scale: 8.0 +2023-03-21 11:35:16,644 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6023, 4.1403, 4.1842, 4.2271, 4.2453, 4.1576, 4.4361, 3.9344], + device='cuda:1'), covar=tensor([0.0146, 0.0158, 0.0121, 0.0172, 0.0380, 0.0112, 0.0145, 0.0183], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0108, 0.0110, 0.0094, 0.0187, 0.0114, 0.0112, 0.0121], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 11:35:26,688 INFO [train.py:901] (1/2) Epoch 44, batch 2200, loss[loss=0.138, simple_loss=0.2229, pruned_loss=0.0266, over 7260.00 frames. ], tot_loss[loss=0.1289, simple_loss=0.2104, pruned_loss=0.02374, over 1441620.09 frames. ], batch size: 64, lr: 3.74e-03, grad_scale: 8.0 +2023-03-21 11:35:33,361 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 11:35:40,755 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-03-21 11:35:41,277 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 +2023-03-21 11:35:41,438 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.824e+02 2.081e+02 2.587e+02 3.989e+02, threshold=4.161e+02, percent-clipped=0.0 +2023-03-21 11:35:46,643 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123670.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:35:53,141 INFO [train.py:901] (1/2) Epoch 44, batch 2250, loss[loss=0.1261, simple_loss=0.2035, pruned_loss=0.02434, over 7283.00 frames. ], tot_loss[loss=0.1287, simple_loss=0.21, pruned_loss=0.02375, over 1441183.60 frames. ], batch size: 70, lr: 3.74e-03, grad_scale: 8.0 +2023-03-21 11:36:06,807 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 11:36:07,279 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 11:36:11,797 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0379, 3.7541, 3.6709, 3.7041, 3.6857, 3.5738, 3.8521, 3.4786], + device='cuda:1'), covar=tensor([0.0133, 0.0167, 0.0153, 0.0186, 0.0412, 0.0129, 0.0170, 0.0205], + device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0107, 0.0109, 0.0094, 0.0186, 0.0114, 0.0112, 0.0121], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 11:36:18,196 INFO [train.py:901] (1/2) Epoch 44, batch 2300, loss[loss=0.09393, simple_loss=0.1535, pruned_loss=0.01717, over 5929.00 frames. ], tot_loss[loss=0.1285, simple_loss=0.2098, pruned_loss=0.02355, over 1441103.21 frames. ], batch size: 25, lr: 3.73e-03, grad_scale: 8.0 +2023-03-21 11:36:18,714 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 11:36:33,462 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 1.843e+02 2.130e+02 2.540e+02 4.098e+02, threshold=4.259e+02, percent-clipped=0.0 +2023-03-21 11:36:44,415 INFO [train.py:901] (1/2) Epoch 44, batch 2350, loss[loss=0.1482, simple_loss=0.2246, pruned_loss=0.03592, over 7350.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2097, pruned_loss=0.02335, over 1439501.96 frames. ], batch size: 73, lr: 3.73e-03, grad_scale: 8.0 +2023-03-21 11:36:45,006 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123784.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:36:51,890 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 +2023-03-21 11:37:06,573 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 11:37:10,571 INFO [train.py:901] (1/2) Epoch 44, batch 2400, loss[loss=0.1372, simple_loss=0.2155, pruned_loss=0.02945, over 7296.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2099, pruned_loss=0.02328, over 1440575.72 frames. ], batch size: 57, lr: 3.73e-03, grad_scale: 8.0 +2023-03-21 11:37:10,857 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-03-21 11:37:12,995 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 11:37:24,366 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 11:37:25,328 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.028e+02 1.694e+02 1.923e+02 2.237e+02 4.292e+02, threshold=3.846e+02, percent-clipped=1.0 +2023-03-21 11:37:27,405 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 11:37:31,101 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123872.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:37:36,349 INFO [train.py:901] (1/2) Epoch 44, batch 2450, loss[loss=0.136, simple_loss=0.2158, pruned_loss=0.02806, over 7283.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2096, pruned_loss=0.02335, over 1439867.50 frames. ], batch size: 70, lr: 3.73e-03, grad_scale: 8.0 +2023-03-21 11:37:37,003 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4934, 4.3592, 3.8714, 3.9939, 3.4701, 2.5256, 2.0452, 4.4897], + device='cuda:1'), covar=tensor([0.0040, 0.0045, 0.0093, 0.0067, 0.0131, 0.0525, 0.0584, 0.0041], + device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0094, 0.0116, 0.0098, 0.0133, 0.0136, 0.0130, 0.0106], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 11:37:54,162 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 11:37:59,381 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.9857, 2.3795, 1.8113, 2.8059, 2.6623, 2.8651, 2.4233, 2.4998], + device='cuda:1'), covar=tensor([0.2803, 0.1478, 0.4644, 0.0888, 0.0454, 0.0417, 0.0450, 0.0474], + device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0231, 0.0244, 0.0260, 0.0202, 0.0205, 0.0220, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:38:02,813 INFO [train.py:901] (1/2) Epoch 44, batch 2500, loss[loss=0.1255, simple_loss=0.2083, pruned_loss=0.02129, over 7323.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2094, pruned_loss=0.02342, over 1440257.39 frames. ], batch size: 83, lr: 3.73e-03, grad_scale: 8.0 +2023-03-21 11:38:02,963 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123933.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:38:05,235 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-03-21 11:38:16,734 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.179e+02 1.717e+02 1.944e+02 2.434e+02 4.068e+02, threshold=3.888e+02, percent-clipped=1.0 +2023-03-21 11:38:19,708 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 11:38:21,342 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123970.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:38:24,913 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.2845, 3.4137, 2.2499, 3.7791, 3.0022, 3.1816, 1.6760, 2.3778], + device='cuda:1'), covar=tensor([0.0499, 0.0707, 0.3338, 0.0603, 0.0542, 0.0717, 0.4513, 0.2251], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0257, 0.0275, 0.0270, 0.0267, 0.0266, 0.0229, 0.0258], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:38:27,698 INFO [train.py:901] (1/2) Epoch 44, batch 2550, loss[loss=0.1463, simple_loss=0.2297, pruned_loss=0.03149, over 6847.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2096, pruned_loss=0.02328, over 1442354.08 frames. ], batch size: 107, lr: 3.73e-03, grad_scale: 8.0 +2023-03-21 11:38:50,875 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=124018.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:38:50,966 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5406, 1.8261, 1.4703, 1.7032, 1.8747, 1.7765, 1.6454, 1.2987], + device='cuda:1'), covar=tensor([0.0219, 0.0187, 0.0331, 0.0191, 0.0169, 0.0145, 0.0175, 0.0226], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0039, 0.0037, 0.0040, 0.0037, 0.0036, 0.0039, 0.0047], + device='cuda:1'), out_proj_covar=tensor([4.4637e-05, 4.3514e-05, 4.2009e-05, 4.3911e-05, 4.0999e-05, 4.0136e-05, + 4.3143e-05, 5.1681e-05], device='cuda:1') +2023-03-21 11:38:51,244 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-03-21 11:38:58,385 INFO [train.py:901] (1/2) Epoch 44, batch 2600, loss[loss=0.123, simple_loss=0.2132, pruned_loss=0.01646, over 7269.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2102, pruned_loss=0.02326, over 1442825.80 frames. ], batch size: 57, lr: 3.73e-03, grad_scale: 8.0 +2023-03-21 11:39:00,113 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124036.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:39:12,320 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.198e+02 1.686e+02 1.984e+02 2.417e+02 5.230e+02, threshold=3.969e+02, percent-clipped=2.0 +2023-03-21 11:39:12,906 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124062.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:39:22,973 INFO [train.py:901] (1/2) Epoch 44, batch 2650, loss[loss=0.1247, simple_loss=0.2049, pruned_loss=0.02225, over 7259.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2102, pruned_loss=0.02307, over 1443646.02 frames. ], batch size: 52, lr: 3.73e-03, grad_scale: 8.0 +2023-03-21 11:39:23,559 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124084.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:39:27,561 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2272, 3.6028, 4.2609, 4.3022, 4.3424, 4.1934, 4.3877, 4.2771], + device='cuda:1'), covar=tensor([0.0029, 0.0113, 0.0030, 0.0028, 0.0026, 0.0034, 0.0025, 0.0040], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0073, 0.0061, 0.0059, 0.0056, 0.0062, 0.0050, 0.0081], + device='cuda:1'), out_proj_covar=tensor([8.4502e-05, 1.4608e-04, 1.0752e-04, 1.0006e-04, 9.3555e-05, 1.0619e-04, + 9.3685e-05, 1.4665e-04], device='cuda:1') +2023-03-21 11:39:30,138 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124097.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:39:38,747 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3910, 3.0020, 3.2179, 3.3289, 3.0741, 2.9101, 3.3584, 2.4048], + device='cuda:1'), covar=tensor([0.0704, 0.0602, 0.0833, 0.0754, 0.0809, 0.1251, 0.0832, 0.2683], + device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0336, 0.0271, 0.0353, 0.0282, 0.0286, 0.0348, 0.0239], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:39:43,218 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124123.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:39:44,857 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 +2023-03-21 11:39:47,604 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=124132.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:39:48,031 INFO [train.py:901] (1/2) Epoch 44, batch 2700, loss[loss=0.1282, simple_loss=0.2085, pruned_loss=0.02394, over 7313.00 frames. ], tot_loss[loss=0.1284, simple_loss=0.2105, pruned_loss=0.02314, over 1444788.50 frames. ], batch size: 83, lr: 3.73e-03, grad_scale: 8.0 +2023-03-21 11:40:01,924 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.189e+02 1.704e+02 1.876e+02 2.261e+02 3.990e+02, threshold=3.752e+02, percent-clipped=1.0 +2023-03-21 11:40:12,779 INFO [train.py:901] (1/2) Epoch 44, batch 2750, loss[loss=0.1299, simple_loss=0.2108, pruned_loss=0.0245, over 7310.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2103, pruned_loss=0.02299, over 1444869.38 frames. ], batch size: 80, lr: 3.73e-03, grad_scale: 8.0 +2023-03-21 11:40:29,518 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0240, 3.0874, 3.9493, 4.0433, 4.1245, 3.9894, 4.1211, 4.0033], + device='cuda:1'), covar=tensor([0.0031, 0.0147, 0.0037, 0.0031, 0.0028, 0.0034, 0.0035, 0.0045], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0074, 0.0061, 0.0059, 0.0056, 0.0062, 0.0050, 0.0081], + device='cuda:1'), out_proj_covar=tensor([8.4928e-05, 1.4658e-04, 1.0770e-04, 1.0044e-04, 9.3352e-05, 1.0615e-04, + 9.3924e-05, 1.4688e-04], device='cuda:1') +2023-03-21 11:40:31,112 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-03-21 11:40:35,765 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124228.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:40:37,873 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8344, 1.7045, 2.2512, 2.3885, 2.2097, 2.3295, 2.1364, 2.4117], + device='cuda:1'), covar=tensor([0.2731, 0.4611, 0.1216, 0.1511, 0.1454, 0.1893, 0.2641, 0.2154], + device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0082, 0.0075, 0.0069, 0.0068, 0.0067, 0.0108, 0.0069], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:40:38,227 INFO [train.py:901] (1/2) Epoch 44, batch 2800, loss[loss=0.1206, simple_loss=0.1991, pruned_loss=0.02112, over 7264.00 frames. ], tot_loss[loss=0.1279, simple_loss=0.2099, pruned_loss=0.02296, over 1443766.93 frames. ], batch size: 47, lr: 3.73e-03, grad_scale: 8.0 +2023-03-21 11:41:02,383 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 11:41:02,710 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 11:41:02,734 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 11:41:02,749 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 11:41:02,917 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 11:41:02,932 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 11:41:02,943 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 11:41:03,038 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 11:41:03,111 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 11:41:03,125 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 11:41:03,142 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 11:41:03,299 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 11:41:03,305 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 11:41:03,312 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 11:41:03,749 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 11:41:03,752 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 11:41:03,808 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 11:41:10,378 INFO [train.py:901] (1/2) Epoch 45, batch 0, loss[loss=0.1235, simple_loss=0.2073, pruned_loss=0.01989, over 7322.00 frames. ], tot_loss[loss=0.1235, simple_loss=0.2073, pruned_loss=0.01989, over 7322.00 frames. ], batch size: 59, lr: 3.69e-03, grad_scale: 8.0 +2023-03-21 11:41:10,378 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 11:41:36,461 INFO [train.py:935] (1/2) Epoch 45, validation: loss=0.1662, simple_loss=0.2586, pruned_loss=0.03694, over 1622729.00 frames. +2023-03-21 11:41:36,462 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 11:41:39,081 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.245e+02 1.746e+02 2.054e+02 2.421e+02 4.265e+02, threshold=4.108e+02, percent-clipped=2.0 +2023-03-21 11:41:43,532 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 11:41:44,690 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124272.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:41:53,502 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 11:41:55,861 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-21 11:42:00,553 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 11:42:01,979 INFO [train.py:901] (1/2) Epoch 45, batch 50, loss[loss=0.1142, simple_loss=0.2011, pruned_loss=0.01368, over 7301.00 frames. ], tot_loss[loss=0.1275, simple_loss=0.21, pruned_loss=0.0225, over 327844.31 frames. ], batch size: 77, lr: 3.68e-03, grad_scale: 8.0 +2023-03-21 11:42:03,027 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 11:42:05,493 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 11:42:13,075 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2965, 2.2550, 2.6558, 2.3355, 2.3977, 2.3792, 2.2934, 1.8375], + device='cuda:1'), covar=tensor([0.0475, 0.0558, 0.0306, 0.0242, 0.0649, 0.0406, 0.0348, 0.0450], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0042, 0.0043, 0.0042, 0.0039, 0.0039, 0.0045, 0.0045], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 11:42:15,134 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124333.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:42:22,738 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 11:42:23,250 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 11:42:24,043 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124348.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 11:42:28,445 INFO [train.py:901] (1/2) Epoch 45, batch 100, loss[loss=0.1323, simple_loss=0.2143, pruned_loss=0.02515, over 7284.00 frames. ], tot_loss[loss=0.1268, simple_loss=0.2088, pruned_loss=0.02239, over 573768.86 frames. ], batch size: 86, lr: 3.68e-03, grad_scale: 8.0 +2023-03-21 11:42:30,318 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.695e+02 1.936e+02 2.313e+02 3.507e+02, threshold=3.872e+02, percent-clipped=0.0 +2023-03-21 11:42:40,537 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8539, 1.6075, 2.1926, 2.3157, 2.2133, 2.4356, 2.1059, 2.4702], + device='cuda:1'), covar=tensor([0.3683, 0.3365, 0.1959, 0.1040, 0.1224, 0.1594, 0.1735, 0.1938], + device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0082, 0.0075, 0.0068, 0.0068, 0.0067, 0.0107, 0.0069], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:42:45,998 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124392.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:42:52,633 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 +2023-03-21 11:42:53,743 INFO [train.py:901] (1/2) Epoch 45, batch 150, loss[loss=0.1593, simple_loss=0.2448, pruned_loss=0.03692, over 6685.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.2093, pruned_loss=0.02303, over 766395.17 frames. ], batch size: 106, lr: 3.68e-03, grad_scale: 8.0 +2023-03-21 11:42:54,909 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124409.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 11:42:59,353 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124418.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:43:02,567 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-03-21 11:43:07,866 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124432.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:43:20,193 INFO [train.py:901] (1/2) Epoch 45, batch 200, loss[loss=0.122, simple_loss=0.2011, pruned_loss=0.02147, over 7271.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.2094, pruned_loss=0.02302, over 916816.88 frames. ], batch size: 47, lr: 3.68e-03, grad_scale: 8.0 +2023-03-21 11:43:22,217 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.227e+02 1.748e+02 2.057e+02 2.417e+02 4.468e+02, threshold=4.114e+02, percent-clipped=4.0 +2023-03-21 11:43:22,746 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 11:43:26,636 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 11:43:32,685 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 11:43:38,371 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124493.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:43:40,839 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7074, 1.5570, 1.7445, 2.0256, 1.7810, 2.0300, 1.5515, 2.1412], + device='cuda:1'), covar=tensor([0.2155, 0.3871, 0.1781, 0.1249, 0.2848, 0.2375, 0.1958, 0.1698], + device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0081, 0.0075, 0.0068, 0.0067, 0.0066, 0.0106, 0.0068], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:43:45,279 INFO [train.py:901] (1/2) Epoch 45, batch 250, loss[loss=0.1358, simple_loss=0.2125, pruned_loss=0.02953, over 7265.00 frames. ], tot_loss[loss=0.1274, simple_loss=0.209, pruned_loss=0.02285, over 1031232.45 frames. ], batch size: 89, lr: 3.68e-03, grad_scale: 8.0 +2023-03-21 11:43:45,787 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 11:43:56,956 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124528.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:44:06,285 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 11:44:11,386 INFO [train.py:901] (1/2) Epoch 45, batch 300, loss[loss=0.1337, simple_loss=0.2206, pruned_loss=0.02343, over 7244.00 frames. ], tot_loss[loss=0.128, simple_loss=0.2101, pruned_loss=0.02296, over 1123807.45 frames. ], batch size: 55, lr: 3.68e-03, grad_scale: 8.0 +2023-03-21 11:44:13,401 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.695e+02 1.987e+02 2.294e+02 4.140e+02, threshold=3.974e+02, percent-clipped=0.0 +2023-03-21 11:44:15,454 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 11:44:21,062 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=124576.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:44:38,054 INFO [train.py:901] (1/2) Epoch 45, batch 350, loss[loss=0.136, simple_loss=0.2154, pruned_loss=0.02832, over 7275.00 frames. ], tot_loss[loss=0.1279, simple_loss=0.2097, pruned_loss=0.02302, over 1196062.01 frames. ], batch size: 52, lr: 3.68e-03, grad_scale: 8.0 +2023-03-21 11:44:48,619 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 11:44:48,679 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124628.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:44:53,753 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124638.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:45:03,204 INFO [train.py:901] (1/2) Epoch 45, batch 400, loss[loss=0.1202, simple_loss=0.203, pruned_loss=0.01868, over 7327.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.2095, pruned_loss=0.02292, over 1250815.99 frames. ], batch size: 75, lr: 3.68e-03, grad_scale: 8.0 +2023-03-21 11:45:05,237 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.772e+02 2.009e+02 2.324e+02 6.365e+02, threshold=4.017e+02, percent-clipped=3.0 +2023-03-21 11:45:06,409 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124663.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:45:22,320 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124692.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:45:24,383 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5202, 1.8285, 1.5048, 1.6934, 1.8366, 1.9248, 1.7109, 1.4909], + device='cuda:1'), covar=tensor([0.0183, 0.0261, 0.0286, 0.0241, 0.0158, 0.0132, 0.0170, 0.0211], + device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0039, 0.0037, 0.0039, 0.0037, 0.0036, 0.0039, 0.0047], + device='cuda:1'), out_proj_covar=tensor([4.4492e-05, 4.3409e-05, 4.1768e-05, 4.3431e-05, 4.0846e-05, 3.9827e-05, + 4.3214e-05, 5.1603e-05], device='cuda:1') +2023-03-21 11:45:25,861 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124699.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:45:28,304 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124704.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 11:45:29,686 INFO [train.py:901] (1/2) Epoch 45, batch 450, loss[loss=0.1242, simple_loss=0.2059, pruned_loss=0.02122, over 7265.00 frames. ], tot_loss[loss=0.1279, simple_loss=0.2094, pruned_loss=0.02316, over 1293045.38 frames. ], batch size: 64, lr: 3.68e-03, grad_scale: 8.0 +2023-03-21 11:45:31,718 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 11:45:32,152 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 11:45:35,300 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124718.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:45:38,343 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124724.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:45:46,310 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=124740.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:45:50,411 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7947, 2.9924, 3.8593, 3.6810, 3.9950, 3.9178, 3.8797, 3.7029], + device='cuda:1'), covar=tensor([0.0056, 0.0207, 0.0047, 0.0066, 0.0044, 0.0047, 0.0061, 0.0082], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0073, 0.0061, 0.0059, 0.0056, 0.0062, 0.0051, 0.0082], + device='cuda:1'), out_proj_covar=tensor([8.5531e-05, 1.4541e-04, 1.0664e-04, 1.0050e-04, 9.3778e-05, 1.0558e-04, + 9.4791e-05, 1.4751e-04], device='cuda:1') +2023-03-21 11:45:54,793 INFO [train.py:901] (1/2) Epoch 45, batch 500, loss[loss=0.1024, simple_loss=0.1752, pruned_loss=0.01485, over 7155.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.2091, pruned_loss=0.02314, over 1326607.53 frames. ], batch size: 39, lr: 3.68e-03, grad_scale: 8.0 +2023-03-21 11:45:56,803 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.234e+02 1.743e+02 1.901e+02 2.274e+02 3.978e+02, threshold=3.801e+02, percent-clipped=0.0 +2023-03-21 11:46:00,037 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=124766.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:46:06,203 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 11:46:07,732 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 11:46:08,753 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 11:46:10,770 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 11:46:11,905 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124788.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:46:14,926 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5611, 3.0222, 2.1551, 3.2874, 3.2880, 3.5810, 3.0862, 2.7567], + device='cuda:1'), covar=tensor([0.2409, 0.1071, 0.4538, 0.0822, 0.0361, 0.0430, 0.0558, 0.0537], + device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0232, 0.0246, 0.0260, 0.0202, 0.0205, 0.0222, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:46:15,288 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 11:46:21,538 INFO [train.py:901] (1/2) Epoch 45, batch 550, loss[loss=0.1335, simple_loss=0.218, pruned_loss=0.02443, over 7302.00 frames. ], tot_loss[loss=0.1275, simple_loss=0.209, pruned_loss=0.02299, over 1353784.97 frames. ], batch size: 80, lr: 3.68e-03, grad_scale: 8.0 +2023-03-21 11:46:26,602 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 11:46:28,263 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124820.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:46:29,778 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2946, 2.2283, 2.6145, 2.3246, 2.3632, 2.4325, 2.2640, 1.8654], + device='cuda:1'), covar=tensor([0.0496, 0.0559, 0.0326, 0.0295, 0.0493, 0.0529, 0.0375, 0.0358], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0041, 0.0042, 0.0041, 0.0038, 0.0038, 0.0044, 0.0044], + device='cuda:1'), out_proj_covar=tensor([1.0585e-04, 1.0529e-04, 1.0535e-04, 1.0401e-04, 1.0082e-04, 9.9858e-05, + 1.1052e-04, 1.1090e-04], device='cuda:1') +2023-03-21 11:46:34,836 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 11:46:38,238 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 11:46:44,967 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 11:46:47,465 INFO [train.py:901] (1/2) Epoch 45, batch 600, loss[loss=0.1312, simple_loss=0.214, pruned_loss=0.02422, over 7338.00 frames. ], tot_loss[loss=0.1272, simple_loss=0.2086, pruned_loss=0.02286, over 1372142.62 frames. ], batch size: 54, lr: 3.68e-03, grad_scale: 8.0 +2023-03-21 11:46:50,048 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.309e+02 1.678e+02 1.954e+02 2.245e+02 3.670e+02, threshold=3.907e+02, percent-clipped=0.0 +2023-03-21 11:47:00,355 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124881.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:47:01,738 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 11:47:10,732 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 11:47:13,265 INFO [train.py:901] (1/2) Epoch 45, batch 650, loss[loss=0.1237, simple_loss=0.2109, pruned_loss=0.01828, over 7285.00 frames. ], tot_loss[loss=0.1272, simple_loss=0.2088, pruned_loss=0.0228, over 1387284.57 frames. ], batch size: 68, lr: 3.68e-03, grad_scale: 8.0 +2023-03-21 11:47:17,994 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1776, 2.4931, 1.9355, 3.1231, 2.7118, 3.2614, 2.7521, 2.7165], + device='cuda:1'), covar=tensor([0.2529, 0.1333, 0.4513, 0.0794, 0.0363, 0.0388, 0.0508, 0.0428], + device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0231, 0.0245, 0.0259, 0.0201, 0.0204, 0.0222, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:47:23,977 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124928.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:47:28,465 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 11:47:37,600 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 11:47:37,920 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 +2023-03-21 11:47:39,633 INFO [train.py:901] (1/2) Epoch 45, batch 700, loss[loss=0.1143, simple_loss=0.1956, pruned_loss=0.0165, over 7361.00 frames. ], tot_loss[loss=0.1273, simple_loss=0.2089, pruned_loss=0.02283, over 1399408.48 frames. ], batch size: 73, lr: 3.67e-03, grad_scale: 8.0 +2023-03-21 11:47:41,668 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.255e+02 1.758e+02 2.104e+02 2.521e+02 4.438e+02, threshold=4.207e+02, percent-clipped=2.0 +2023-03-21 11:47:49,184 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=124976.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:47:58,300 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124994.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:48:01,324 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 11:48:02,165 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 11:48:03,736 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125004.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 11:48:05,256 INFO [train.py:901] (1/2) Epoch 45, batch 750, loss[loss=0.1333, simple_loss=0.2096, pruned_loss=0.02846, over 7282.00 frames. ], tot_loss[loss=0.1274, simple_loss=0.2086, pruned_loss=0.02311, over 1405824.96 frames. ], batch size: 68, lr: 3.67e-03, grad_scale: 8.0 +2023-03-21 11:48:08,407 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1851, 2.4049, 1.8393, 2.9605, 2.7339, 3.0416, 2.7918, 2.8172], + device='cuda:1'), covar=tensor([0.2257, 0.1237, 0.4210, 0.0894, 0.0407, 0.0390, 0.0499, 0.0441], + device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0232, 0.0245, 0.0260, 0.0202, 0.0204, 0.0221, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:48:08,532 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-03-21 11:48:11,400 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125019.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:48:15,992 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 11:48:21,593 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 11:48:28,149 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 11:48:29,174 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 11:48:29,211 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=125052.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 11:48:31,665 INFO [train.py:901] (1/2) Epoch 45, batch 800, loss[loss=0.09654, simple_loss=0.164, pruned_loss=0.01452, over 6539.00 frames. ], tot_loss[loss=0.1265, simple_loss=0.2077, pruned_loss=0.02271, over 1411416.22 frames. ], batch size: 28, lr: 3.67e-03, grad_scale: 8.0 +2023-03-21 11:48:33,690 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.703e+02 1.941e+02 2.250e+02 3.887e+02, threshold=3.881e+02, percent-clipped=0.0 +2023-03-21 11:48:40,406 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 11:48:47,465 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125088.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:48:56,885 INFO [train.py:901] (1/2) Epoch 45, batch 850, loss[loss=0.122, simple_loss=0.2093, pruned_loss=0.01732, over 7294.00 frames. ], tot_loss[loss=0.1268, simple_loss=0.208, pruned_loss=0.02279, over 1419726.69 frames. ], batch size: 66, lr: 3.67e-03, grad_scale: 8.0 +2023-03-21 11:48:58,977 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 11:48:58,994 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 11:49:04,619 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 11:49:08,151 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 11:49:08,792 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125128.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:49:10,313 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0498, 2.0527, 2.2696, 2.1032, 2.1514, 2.1296, 2.0726, 1.5970], + device='cuda:1'), covar=tensor([0.0358, 0.0456, 0.0276, 0.0306, 0.0387, 0.0481, 0.0345, 0.0338], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0041, 0.0042, 0.0041, 0.0038, 0.0039, 0.0045, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 11:49:12,744 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=125136.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:49:13,249 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7249, 4.2580, 4.0324, 4.6231, 4.4148, 4.5937, 4.0226, 4.2716], + device='cuda:1'), covar=tensor([0.0909, 0.2526, 0.2941, 0.1148, 0.1144, 0.1362, 0.0911, 0.1376], + device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0402, 0.0302, 0.0319, 0.0236, 0.0375, 0.0236, 0.0283], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 11:49:23,160 INFO [train.py:901] (1/2) Epoch 45, batch 900, loss[loss=0.1337, simple_loss=0.2069, pruned_loss=0.03028, over 7245.00 frames. ], tot_loss[loss=0.1271, simple_loss=0.2085, pruned_loss=0.02286, over 1424524.14 frames. ], batch size: 45, lr: 3.67e-03, grad_scale: 8.0 +2023-03-21 11:49:25,153 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.791e+02 2.042e+02 2.328e+02 6.599e+02, threshold=4.084e+02, percent-clipped=1.0 +2023-03-21 11:49:32,748 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125176.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:49:39,369 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125189.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:49:47,738 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 11:49:49,225 INFO [train.py:901] (1/2) Epoch 45, batch 950, loss[loss=0.1369, simple_loss=0.2218, pruned_loss=0.02598, over 7333.00 frames. ], tot_loss[loss=0.1263, simple_loss=0.2076, pruned_loss=0.02253, over 1422891.01 frames. ], batch size: 61, lr: 3.67e-03, grad_scale: 8.0 +2023-03-21 11:50:11,377 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 11:50:14,959 INFO [train.py:901] (1/2) Epoch 45, batch 1000, loss[loss=0.1188, simple_loss=0.2074, pruned_loss=0.01515, over 7256.00 frames. ], tot_loss[loss=0.1266, simple_loss=0.208, pruned_loss=0.02264, over 1424644.20 frames. ], batch size: 89, lr: 3.67e-03, grad_scale: 16.0 +2023-03-21 11:50:16,955 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.326e+02 1.650e+02 2.037e+02 2.329e+02 4.165e+02, threshold=4.073e+02, percent-clipped=1.0 +2023-03-21 11:50:32,583 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 11:50:34,234 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125294.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:50:41,358 INFO [train.py:901] (1/2) Epoch 45, batch 1050, loss[loss=0.1438, simple_loss=0.2218, pruned_loss=0.03284, over 7255.00 frames. ], tot_loss[loss=0.1269, simple_loss=0.2083, pruned_loss=0.02274, over 1427593.40 frames. ], batch size: 64, lr: 3.67e-03, grad_scale: 16.0 +2023-03-21 11:50:47,373 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7466, 5.2080, 5.2828, 5.2126, 5.0516, 4.7897, 5.3103, 5.1496], + device='cuda:1'), covar=tensor([0.0428, 0.0359, 0.0331, 0.0450, 0.0310, 0.0394, 0.0290, 0.0361], + device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0271, 0.0208, 0.0208, 0.0160, 0.0235, 0.0215, 0.0152], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:50:47,400 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125319.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:50:54,317 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 11:50:58,352 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 11:50:58,879 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=125342.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:51:04,795 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 +2023-03-21 11:51:06,094 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125356.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:51:06,440 INFO [train.py:901] (1/2) Epoch 45, batch 1100, loss[loss=0.1241, simple_loss=0.2043, pruned_loss=0.02198, over 7317.00 frames. ], tot_loss[loss=0.1271, simple_loss=0.2083, pruned_loss=0.023, over 1431088.19 frames. ], batch size: 61, lr: 3.67e-03, grad_scale: 16.0 +2023-03-21 11:51:08,425 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.764e+02 2.014e+02 2.431e+02 3.712e+02, threshold=4.027e+02, percent-clipped=0.0 +2023-03-21 11:51:11,502 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=125367.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:51:16,834 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125376.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:51:20,046 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=2.86 vs. limit=5.0 +2023-03-21 11:51:28,852 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 11:51:29,424 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 11:51:32,844 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3423, 2.3776, 2.5380, 2.3428, 2.4157, 2.2964, 2.2108, 1.8031], + device='cuda:1'), covar=tensor([0.0588, 0.0411, 0.0485, 0.0301, 0.0613, 0.0481, 0.0446, 0.0459], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0041, 0.0042, 0.0041, 0.0038, 0.0039, 0.0045, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 11:51:33,208 INFO [train.py:901] (1/2) Epoch 45, batch 1150, loss[loss=0.1545, simple_loss=0.2336, pruned_loss=0.03765, over 7326.00 frames. ], tot_loss[loss=0.1273, simple_loss=0.2084, pruned_loss=0.02315, over 1432752.11 frames. ], batch size: 59, lr: 3.67e-03, grad_scale: 16.0 +2023-03-21 11:51:38,459 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125417.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 11:51:40,396 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125421.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:51:42,320 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 11:51:42,779 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 11:51:48,378 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125437.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:51:50,020 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 +2023-03-21 11:51:58,987 INFO [train.py:901] (1/2) Epoch 45, batch 1200, loss[loss=0.1157, simple_loss=0.1972, pruned_loss=0.01714, over 7231.00 frames. ], tot_loss[loss=0.1271, simple_loss=0.2082, pruned_loss=0.023, over 1432132.04 frames. ], batch size: 45, lr: 3.67e-03, grad_scale: 16.0 +2023-03-21 11:52:00,946 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.253e+02 1.727e+02 2.078e+02 2.554e+02 4.178e+02, threshold=4.157e+02, percent-clipped=1.0 +2023-03-21 11:52:06,713 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9601, 3.8884, 3.1414, 4.2317, 3.6317, 3.9422, 2.2715, 3.1632], + device='cuda:1'), covar=tensor([0.0538, 0.0955, 0.2265, 0.0599, 0.0535, 0.0866, 0.3524, 0.1733], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0256, 0.0275, 0.0267, 0.0265, 0.0262, 0.0228, 0.0255], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:52:08,686 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125476.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:52:12,319 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125482.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:52:13,250 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125484.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:52:14,484 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-03-21 11:52:15,205 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 11:52:15,276 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2102, 4.6850, 4.7488, 4.6539, 4.6406, 4.2871, 4.7853, 4.6547], + device='cuda:1'), covar=tensor([0.0498, 0.0425, 0.0425, 0.0662, 0.0373, 0.0433, 0.0377, 0.0468], + device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0271, 0.0209, 0.0209, 0.0161, 0.0235, 0.0216, 0.0152], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:52:20,039 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 +2023-03-21 11:52:24,730 INFO [train.py:901] (1/2) Epoch 45, batch 1250, loss[loss=0.132, simple_loss=0.2056, pruned_loss=0.02915, over 7222.00 frames. ], tot_loss[loss=0.1275, simple_loss=0.2086, pruned_loss=0.02314, over 1435588.73 frames. ], batch size: 45, lr: 3.67e-03, grad_scale: 16.0 +2023-03-21 11:52:25,988 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 +2023-03-21 11:52:33,326 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=125524.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:52:38,846 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 11:52:43,437 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 11:52:44,465 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 11:52:50,562 INFO [train.py:901] (1/2) Epoch 45, batch 1300, loss[loss=0.106, simple_loss=0.1874, pruned_loss=0.0123, over 7333.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.209, pruned_loss=0.02323, over 1438425.53 frames. ], batch size: 44, lr: 3.67e-03, grad_scale: 16.0 +2023-03-21 11:52:52,513 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 1.582e+02 2.005e+02 2.378e+02 3.739e+02, threshold=4.009e+02, percent-clipped=0.0 +2023-03-21 11:53:06,660 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 11:53:08,677 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 11:53:12,158 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 11:53:14,629 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125603.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:53:16,430 INFO [train.py:901] (1/2) Epoch 45, batch 1350, loss[loss=0.1418, simple_loss=0.2219, pruned_loss=0.03085, over 7261.00 frames. ], tot_loss[loss=0.1286, simple_loss=0.2099, pruned_loss=0.02362, over 1441538.53 frames. ], batch size: 47, lr: 3.67e-03, grad_scale: 8.0 +2023-03-21 11:53:21,065 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5276, 2.2984, 2.4690, 3.7246, 2.0575, 3.4198, 1.5076, 3.3778], + device='cuda:1'), covar=tensor([0.0188, 0.1592, 0.1911, 0.0230, 0.3919, 0.0326, 0.1300, 0.0478], + device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0245, 0.0257, 0.0210, 0.0249, 0.0218, 0.0221, 0.0229], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:53:22,932 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 11:53:42,803 INFO [train.py:901] (1/2) Epoch 45, batch 1400, loss[loss=0.1226, simple_loss=0.2039, pruned_loss=0.02065, over 7259.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.2099, pruned_loss=0.0233, over 1442014.13 frames. ], batch size: 89, lr: 3.66e-03, grad_scale: 8.0 +2023-03-21 11:53:45,262 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.279e+02 1.672e+02 2.023e+02 2.311e+02 3.601e+02, threshold=4.045e+02, percent-clipped=0.0 +2023-03-21 11:53:46,462 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125664.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:53:55,413 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 11:54:07,911 INFO [train.py:901] (1/2) Epoch 45, batch 1450, loss[loss=0.09676, simple_loss=0.1616, pruned_loss=0.01598, over 6237.00 frames. ], tot_loss[loss=0.1278, simple_loss=0.2093, pruned_loss=0.02316, over 1440426.79 frames. ], batch size: 26, lr: 3.66e-03, grad_scale: 8.0 +2023-03-21 11:54:11,205 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125712.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 11:54:17,661 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125725.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:54:19,656 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9851, 3.3859, 3.0465, 3.2674, 3.2594, 2.8312, 3.2912, 3.1140], + device='cuda:1'), covar=tensor([0.0615, 0.0942, 0.1341, 0.1124, 0.1006, 0.1119, 0.0928, 0.0965], + device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0061, 0.0070, 0.0061, 0.0058, 0.0065, 0.0058, 0.0055], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:54:20,026 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 11:54:21,124 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125732.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:54:30,682 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.2965, 4.7301, 4.6005, 5.1829, 5.0171, 5.0936, 4.5339, 4.7903], + device='cuda:1'), covar=tensor([0.0824, 0.2406, 0.2176, 0.1045, 0.0929, 0.1189, 0.0918, 0.1079], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0398, 0.0299, 0.0316, 0.0236, 0.0374, 0.0235, 0.0281], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 11:54:34,095 INFO [train.py:901] (1/2) Epoch 45, batch 1500, loss[loss=0.1266, simple_loss=0.2033, pruned_loss=0.02491, over 7267.00 frames. ], tot_loss[loss=0.128, simple_loss=0.2093, pruned_loss=0.02334, over 1439804.99 frames. ], batch size: 52, lr: 3.66e-03, grad_scale: 8.0 +2023-03-21 11:54:36,665 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.334e+02 1.679e+02 1.970e+02 2.341e+02 4.100e+02, threshold=3.939e+02, percent-clipped=1.0 +2023-03-21 11:54:36,709 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 11:54:40,276 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9545, 3.3269, 3.9025, 3.8852, 4.0523, 4.0168, 4.1502, 3.9159], + device='cuda:1'), covar=tensor([0.0034, 0.0116, 0.0032, 0.0032, 0.0025, 0.0028, 0.0031, 0.0045], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0074, 0.0061, 0.0060, 0.0057, 0.0062, 0.0050, 0.0082], + device='cuda:1'), out_proj_covar=tensor([8.5973e-05, 1.4610e-04, 1.0790e-04, 1.0023e-04, 9.3968e-05, 1.0606e-04, + 9.4101e-05, 1.4865e-04], device='cuda:1') +2023-03-21 11:54:44,244 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125777.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:54:47,777 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125784.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:54:48,793 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125786.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:54:52,791 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4935, 5.0176, 5.1394, 5.0721, 4.8956, 4.6182, 5.1723, 4.9750], + device='cuda:1'), covar=tensor([0.0509, 0.0379, 0.0378, 0.0455, 0.0358, 0.0371, 0.0284, 0.0432], + device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0271, 0.0209, 0.0207, 0.0161, 0.0235, 0.0216, 0.0152], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:54:59,484 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-03-21 11:55:00,107 INFO [train.py:901] (1/2) Epoch 45, batch 1550, loss[loss=0.1269, simple_loss=0.2116, pruned_loss=0.02112, over 7112.00 frames. ], tot_loss[loss=0.1281, simple_loss=0.2092, pruned_loss=0.02351, over 1440178.98 frames. ], batch size: 98, lr: 3.66e-03, grad_scale: 8.0 +2023-03-21 11:55:01,649 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 11:55:05,383 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 +2023-03-21 11:55:13,185 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=125832.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:55:18,369 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6324, 1.4126, 1.8441, 2.0220, 1.8227, 2.1973, 1.5702, 2.1012], + device='cuda:1'), covar=tensor([0.3267, 0.3810, 0.1556, 0.1768, 0.3563, 0.1038, 0.2122, 0.1911], + device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0082, 0.0076, 0.0069, 0.0068, 0.0066, 0.0108, 0.0069], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:55:25,718 INFO [train.py:901] (1/2) Epoch 45, batch 1600, loss[loss=0.1124, simple_loss=0.191, pruned_loss=0.01696, over 7167.00 frames. ], tot_loss[loss=0.1272, simple_loss=0.2083, pruned_loss=0.02306, over 1439698.48 frames. ], batch size: 41, lr: 3.66e-03, grad_scale: 8.0 +2023-03-21 11:55:28,119 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.320e+02 1.720e+02 2.021e+02 2.344e+02 4.247e+02, threshold=4.043e+02, percent-clipped=1.0 +2023-03-21 11:55:32,315 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 11:55:32,852 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 11:55:36,079 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8602, 1.5015, 2.2829, 2.3915, 2.2555, 2.5095, 2.1565, 2.3332], + device='cuda:1'), covar=tensor([0.4125, 0.4945, 0.1637, 0.2546, 0.1929, 0.1550, 0.2419, 0.3166], + device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0082, 0.0077, 0.0069, 0.0069, 0.0067, 0.0108, 0.0069], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:55:36,481 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 11:55:46,614 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 11:55:50,605 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 11:55:51,559 INFO [train.py:901] (1/2) Epoch 45, batch 1650, loss[loss=0.1325, simple_loss=0.2103, pruned_loss=0.02737, over 7273.00 frames. ], tot_loss[loss=0.1273, simple_loss=0.2084, pruned_loss=0.02313, over 1439587.12 frames. ], batch size: 70, lr: 3.66e-03, grad_scale: 8.0 +2023-03-21 11:55:59,946 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 11:56:06,190 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9249, 3.1429, 2.2716, 3.3866, 2.4673, 3.0380, 1.5015, 2.2466], + device='cuda:1'), covar=tensor([0.0540, 0.1006, 0.3316, 0.0802, 0.0637, 0.0789, 0.4325, 0.2343], + device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0257, 0.0275, 0.0268, 0.0266, 0.0262, 0.0229, 0.0258], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:56:16,512 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 11:56:17,523 INFO [train.py:901] (1/2) Epoch 45, batch 1700, loss[loss=0.1259, simple_loss=0.2057, pruned_loss=0.02305, over 7284.00 frames. ], tot_loss[loss=0.1271, simple_loss=0.2081, pruned_loss=0.02299, over 1442268.37 frames. ], batch size: 70, lr: 3.66e-03, grad_scale: 8.0 +2023-03-21 11:56:18,614 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125959.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:56:20,106 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.100e+02 1.725e+02 1.970e+02 2.452e+02 4.185e+02, threshold=3.941e+02, percent-clipped=1.0 +2023-03-21 11:56:20,677 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 11:56:26,885 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8775, 3.8462, 2.7466, 3.4726, 2.8434, 2.1536, 1.7423, 3.9215], + device='cuda:1'), covar=tensor([0.0062, 0.0081, 0.0248, 0.0100, 0.0248, 0.0666, 0.0757, 0.0066], + device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0096, 0.0119, 0.0101, 0.0137, 0.0138, 0.0130, 0.0108], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 11:56:31,402 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 11:56:36,305 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0204, 3.3422, 2.8565, 3.0522, 3.1670, 2.8714, 3.2566, 3.0900], + device='cuda:1'), covar=tensor([0.0624, 0.0510, 0.0944, 0.1397, 0.1233, 0.0777, 0.0668, 0.1322], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0061, 0.0070, 0.0061, 0.0058, 0.0065, 0.0059, 0.0055], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:56:39,362 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125998.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:56:44,685 INFO [train.py:901] (1/2) Epoch 45, batch 1750, loss[loss=0.1242, simple_loss=0.2066, pruned_loss=0.02083, over 7254.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.2088, pruned_loss=0.02332, over 1440295.24 frames. ], batch size: 52, lr: 3.66e-03, grad_scale: 8.0 +2023-03-21 11:56:47,334 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126012.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 11:56:55,892 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 11:56:56,922 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 11:56:57,549 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126032.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:57:10,379 INFO [train.py:901] (1/2) Epoch 45, batch 1800, loss[loss=0.1213, simple_loss=0.2074, pruned_loss=0.01758, over 7287.00 frames. ], tot_loss[loss=0.1272, simple_loss=0.2085, pruned_loss=0.02289, over 1440217.02 frames. ], batch size: 68, lr: 3.66e-03, grad_scale: 8.0 +2023-03-21 11:57:12,137 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126059.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:57:12,584 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=126060.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:57:13,514 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.242e+02 1.649e+02 1.912e+02 2.281e+02 4.540e+02, threshold=3.824e+02, percent-clipped=2.0 +2023-03-21 11:57:19,186 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 11:57:21,030 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-03-21 11:57:21,046 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 +2023-03-21 11:57:21,349 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126077.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:57:22,825 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=126080.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:57:23,384 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126081.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:57:27,657 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5717, 2.2930, 2.4781, 3.6540, 2.0102, 3.3501, 1.4530, 3.3604], + device='cuda:1'), covar=tensor([0.0240, 0.1670, 0.2163, 0.0270, 0.4096, 0.0308, 0.1540, 0.0580], + device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0243, 0.0255, 0.0208, 0.0247, 0.0218, 0.0220, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 11:57:34,573 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 11:57:37,147 INFO [train.py:901] (1/2) Epoch 45, batch 1850, loss[loss=0.1282, simple_loss=0.2152, pruned_loss=0.02061, over 7304.00 frames. ], tot_loss[loss=0.1269, simple_loss=0.2081, pruned_loss=0.02286, over 1440352.86 frames. ], batch size: 80, lr: 3.66e-03, grad_scale: 8.0 +2023-03-21 11:57:44,240 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 11:57:46,341 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=126125.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:58:00,762 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 11:58:03,218 INFO [train.py:901] (1/2) Epoch 45, batch 1900, loss[loss=0.1199, simple_loss=0.2045, pruned_loss=0.01762, over 7174.00 frames. ], tot_loss[loss=0.1269, simple_loss=0.2082, pruned_loss=0.02281, over 1441343.41 frames. ], batch size: 39, lr: 3.66e-03, grad_scale: 8.0 +2023-03-21 11:58:05,651 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.674e+02 1.922e+02 2.434e+02 3.525e+02, threshold=3.845e+02, percent-clipped=0.0 +2023-03-21 11:58:25,166 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 11:58:29,528 INFO [train.py:901] (1/2) Epoch 45, batch 1950, loss[loss=0.1189, simple_loss=0.1963, pruned_loss=0.02077, over 7149.00 frames. ], tot_loss[loss=0.1271, simple_loss=0.2084, pruned_loss=0.02288, over 1440617.89 frames. ], batch size: 41, lr: 3.66e-03, grad_scale: 8.0 +2023-03-21 11:58:36,612 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 11:58:38,529 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 +2023-03-21 11:58:41,762 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 11:58:42,770 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 11:58:56,035 INFO [train.py:901] (1/2) Epoch 45, batch 2000, loss[loss=0.1215, simple_loss=0.2071, pruned_loss=0.0179, over 7321.00 frames. ], tot_loss[loss=0.1271, simple_loss=0.2083, pruned_loss=0.02291, over 1441336.96 frames. ], batch size: 80, lr: 3.66e-03, grad_scale: 8.0 +2023-03-21 11:58:57,190 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126259.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:58:58,547 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.249e+02 1.755e+02 1.955e+02 2.409e+02 3.875e+02, threshold=3.909e+02, percent-clipped=1.0 +2023-03-21 11:58:58,599 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 11:59:09,352 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 11:59:17,149 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 11:59:21,725 INFO [train.py:901] (1/2) Epoch 45, batch 2050, loss[loss=0.1451, simple_loss=0.2338, pruned_loss=0.02822, over 6725.00 frames. ], tot_loss[loss=0.1274, simple_loss=0.2091, pruned_loss=0.02285, over 1442919.68 frames. ], batch size: 106, lr: 3.66e-03, grad_scale: 8.0 +2023-03-21 11:59:21,790 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=126307.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:59:28,550 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7000, 1.5709, 2.2384, 2.2922, 2.1466, 2.5670, 2.1191, 2.2530], + device='cuda:1'), covar=tensor([0.2943, 0.4822, 0.1707, 0.0846, 0.3588, 0.1342, 0.2485, 0.7788], + device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0082, 0.0076, 0.0069, 0.0069, 0.0066, 0.0109, 0.0069], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 11:59:46,782 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126354.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 11:59:48,241 INFO [train.py:901] (1/2) Epoch 45, batch 2100, loss[loss=0.1259, simple_loss=0.2043, pruned_loss=0.02372, over 7306.00 frames. ], tot_loss[loss=0.1275, simple_loss=0.209, pruned_loss=0.02298, over 1444877.08 frames. ], batch size: 59, lr: 3.65e-03, grad_scale: 8.0 +2023-03-21 11:59:50,655 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.203e+02 1.670e+02 2.047e+02 2.347e+02 4.553e+02, threshold=4.095e+02, percent-clipped=3.0 +2023-03-21 11:59:51,190 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 11:59:54,713 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 11:59:59,986 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126380.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:00:00,488 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126381.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:00:14,578 INFO [train.py:901] (1/2) Epoch 45, batch 2150, loss[loss=0.1087, simple_loss=0.1892, pruned_loss=0.01414, over 7332.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.2093, pruned_loss=0.02304, over 1445996.41 frames. ], batch size: 44, lr: 3.65e-03, grad_scale: 8.0 +2023-03-21 12:00:19,809 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2216, 2.1448, 2.3821, 2.1499, 2.3143, 2.2533, 2.1948, 1.6973], + device='cuda:1'), covar=tensor([0.0459, 0.0470, 0.0404, 0.0399, 0.0538, 0.0557, 0.0363, 0.0375], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0041, 0.0042, 0.0041, 0.0038, 0.0039, 0.0045, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 12:00:26,287 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=126429.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:00:32,479 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126441.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:00:37,003 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126450.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:00:40,382 INFO [train.py:901] (1/2) Epoch 45, batch 2200, loss[loss=0.1279, simple_loss=0.2104, pruned_loss=0.02273, over 7266.00 frames. ], tot_loss[loss=0.1276, simple_loss=0.2093, pruned_loss=0.02298, over 1445380.17 frames. ], batch size: 47, lr: 3.65e-03, grad_scale: 8.0 +2023-03-21 12:00:41,919 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 12:00:42,884 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 9.798e+01 1.679e+02 2.078e+02 2.432e+02 4.559e+02, threshold=4.157e+02, percent-clipped=2.0 +2023-03-21 12:01:06,382 INFO [train.py:901] (1/2) Epoch 45, batch 2250, loss[loss=0.131, simple_loss=0.2124, pruned_loss=0.0248, over 7258.00 frames. ], tot_loss[loss=0.1275, simple_loss=0.209, pruned_loss=0.02296, over 1443621.21 frames. ], batch size: 89, lr: 3.65e-03, grad_scale: 8.0 +2023-03-21 12:01:08,572 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126511.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:01:16,735 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 12:01:17,215 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 12:01:29,450 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 12:01:32,496 INFO [train.py:901] (1/2) Epoch 45, batch 2300, loss[loss=0.1432, simple_loss=0.2352, pruned_loss=0.02555, over 7247.00 frames. ], tot_loss[loss=0.1276, simple_loss=0.2093, pruned_loss=0.02295, over 1444648.20 frames. ], batch size: 55, lr: 3.65e-03, grad_scale: 8.0 +2023-03-21 12:01:34,913 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.261e+02 1.701e+02 2.024e+02 2.404e+02 3.706e+02, threshold=4.049e+02, percent-clipped=0.0 +2023-03-21 12:01:41,297 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-21 12:01:55,790 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-03-21 12:01:59,157 INFO [train.py:901] (1/2) Epoch 45, batch 2350, loss[loss=0.1069, simple_loss=0.1902, pruned_loss=0.01179, over 7348.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.2094, pruned_loss=0.02298, over 1442479.92 frames. ], batch size: 44, lr: 3.65e-03, grad_scale: 8.0 +2023-03-21 12:02:16,241 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 12:02:22,825 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 12:02:22,949 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126654.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:02:24,363 INFO [train.py:901] (1/2) Epoch 45, batch 2400, loss[loss=0.1244, simple_loss=0.2078, pruned_loss=0.02045, over 7247.00 frames. ], tot_loss[loss=0.1279, simple_loss=0.2095, pruned_loss=0.02316, over 1442888.49 frames. ], batch size: 89, lr: 3.65e-03, grad_scale: 8.0 +2023-03-21 12:02:24,510 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1879, 2.3420, 2.7610, 2.1902, 2.4611, 2.3039, 2.2871, 1.9068], + device='cuda:1'), covar=tensor([0.0700, 0.0455, 0.0346, 0.0451, 0.0584, 0.0606, 0.0445, 0.0462], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0042, 0.0042, 0.0042, 0.0039, 0.0039, 0.0045, 0.0045], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 12:02:27,504 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.239e+02 1.819e+02 2.131e+02 2.403e+02 3.692e+02, threshold=4.262e+02, percent-clipped=0.0 +2023-03-21 12:02:34,143 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 12:02:37,180 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 12:02:48,446 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=126702.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:02:50,912 INFO [train.py:901] (1/2) Epoch 45, batch 2450, loss[loss=0.1353, simple_loss=0.2168, pruned_loss=0.0269, over 7301.00 frames. ], tot_loss[loss=0.1278, simple_loss=0.2096, pruned_loss=0.02298, over 1444481.68 frames. ], batch size: 68, lr: 3.65e-03, grad_scale: 8.0 +2023-03-21 12:02:54,509 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3630, 3.5824, 3.4437, 3.4921, 3.3020, 3.4720, 3.8402, 3.8133], + device='cuda:1'), covar=tensor([0.0324, 0.0208, 0.0278, 0.0237, 0.0405, 0.0507, 0.0235, 0.0237], + device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0130, 0.0124, 0.0130, 0.0119, 0.0105, 0.0101, 0.0106], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 12:03:03,197 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 12:03:05,852 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126736.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:03:08,382 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6987, 4.1871, 4.0143, 4.6275, 4.5066, 4.5769, 4.0988, 4.3016], + device='cuda:1'), covar=tensor([0.0960, 0.2521, 0.2503, 0.1096, 0.1024, 0.1268, 0.0962, 0.1219], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0404, 0.0306, 0.0322, 0.0242, 0.0378, 0.0239, 0.0286], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 12:03:10,543 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4865, 3.6783, 2.5679, 3.9239, 3.0432, 3.5722, 1.7792, 2.7163], + device='cuda:1'), covar=tensor([0.0611, 0.0759, 0.2917, 0.0518, 0.0573, 0.0961, 0.4237, 0.2087], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0255, 0.0276, 0.0267, 0.0266, 0.0261, 0.0227, 0.0257], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:03:17,005 INFO [train.py:901] (1/2) Epoch 45, batch 2500, loss[loss=0.1256, simple_loss=0.2056, pruned_loss=0.02281, over 7265.00 frames. ], tot_loss[loss=0.128, simple_loss=0.2099, pruned_loss=0.02299, over 1444266.74 frames. ], batch size: 55, lr: 3.65e-03, grad_scale: 8.0 +2023-03-21 12:03:19,526 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.802e+02 2.146e+02 2.550e+02 5.501e+02, threshold=4.291e+02, percent-clipped=1.0 +2023-03-21 12:03:28,663 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 12:03:36,214 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 +2023-03-21 12:03:42,814 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126806.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:03:43,254 INFO [train.py:901] (1/2) Epoch 45, batch 2550, loss[loss=0.1206, simple_loss=0.202, pruned_loss=0.01961, over 7292.00 frames. ], tot_loss[loss=0.1278, simple_loss=0.21, pruned_loss=0.02287, over 1444756.70 frames. ], batch size: 86, lr: 3.65e-03, grad_scale: 8.0 +2023-03-21 12:03:44,372 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1903, 3.9121, 3.8245, 3.9001, 3.8836, 3.7424, 4.0988, 3.6929], + device='cuda:1'), covar=tensor([0.0166, 0.0190, 0.0149, 0.0186, 0.0403, 0.0142, 0.0139, 0.0193], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0106, 0.0108, 0.0092, 0.0182, 0.0113, 0.0110, 0.0118], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 12:03:47,947 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126816.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:03:50,969 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1629, 4.4001, 4.1484, 4.3019, 4.0121, 4.0935, 4.5405, 4.5725], + device='cuda:1'), covar=tensor([0.0338, 0.0204, 0.0324, 0.0290, 0.0471, 0.0382, 0.0324, 0.0314], + device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0130, 0.0124, 0.0129, 0.0118, 0.0105, 0.0100, 0.0106], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 12:03:58,898 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.15 vs. limit=2.0 +2023-03-21 12:04:09,072 INFO [train.py:901] (1/2) Epoch 45, batch 2600, loss[loss=0.1167, simple_loss=0.2046, pruned_loss=0.01444, over 7245.00 frames. ], tot_loss[loss=0.1279, simple_loss=0.2104, pruned_loss=0.02275, over 1446322.26 frames. ], batch size: 55, lr: 3.65e-03, grad_scale: 8.0 +2023-03-21 12:04:11,549 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.684e+02 1.885e+02 2.330e+02 6.166e+02, threshold=3.769e+02, percent-clipped=1.0 +2023-03-21 12:04:19,254 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126877.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:04:33,295 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 +2023-03-21 12:04:34,047 INFO [train.py:901] (1/2) Epoch 45, batch 2650, loss[loss=0.1362, simple_loss=0.2166, pruned_loss=0.02791, over 7357.00 frames. ], tot_loss[loss=0.1276, simple_loss=0.2099, pruned_loss=0.02266, over 1444725.46 frames. ], batch size: 54, lr: 3.65e-03, grad_scale: 8.0 +2023-03-21 12:04:59,422 INFO [train.py:901] (1/2) Epoch 45, batch 2700, loss[loss=0.1477, simple_loss=0.2347, pruned_loss=0.03035, over 6684.00 frames. ], tot_loss[loss=0.1278, simple_loss=0.2099, pruned_loss=0.02284, over 1442853.33 frames. ], batch size: 106, lr: 3.65e-03, grad_scale: 8.0 +2023-03-21 12:05:01,891 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.267e+02 1.714e+02 2.002e+02 2.426e+02 5.440e+02, threshold=4.003e+02, percent-clipped=3.0 +2023-03-21 12:05:24,667 INFO [train.py:901] (1/2) Epoch 45, batch 2750, loss[loss=0.1286, simple_loss=0.2139, pruned_loss=0.02169, over 7321.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.2099, pruned_loss=0.02275, over 1443374.42 frames. ], batch size: 61, lr: 3.65e-03, grad_scale: 8.0 +2023-03-21 12:05:38,890 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127036.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:05:48,493 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 12:05:49,156 INFO [train.py:901] (1/2) Epoch 45, batch 2800, loss[loss=0.1394, simple_loss=0.2231, pruned_loss=0.02785, over 7260.00 frames. ], tot_loss[loss=0.1283, simple_loss=0.2103, pruned_loss=0.02317, over 1444466.19 frames. ], batch size: 55, lr: 3.64e-03, grad_scale: 8.0 +2023-03-21 12:05:52,003 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.659e+02 1.865e+02 2.207e+02 4.110e+02, threshold=3.730e+02, percent-clipped=1.0 +2023-03-21 12:06:14,862 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 12:06:14,914 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 12:06:14,939 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 12:06:14,955 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 12:06:15,129 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 12:06:15,145 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 12:06:15,155 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 12:06:15,257 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 12:06:15,332 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 12:06:15,346 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 12:06:15,363 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 12:06:15,885 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 12:06:15,891 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 12:06:15,898 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 12:06:16,034 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 12:06:16,036 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 12:06:16,093 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 12:06:21,167 INFO [train.py:901] (1/2) Epoch 46, batch 0, loss[loss=0.1316, simple_loss=0.2189, pruned_loss=0.02212, over 7368.00 frames. ], tot_loss[loss=0.1316, simple_loss=0.2189, pruned_loss=0.02212, over 7368.00 frames. ], batch size: 65, lr: 3.60e-03, grad_scale: 8.0 +2023-03-21 12:06:21,168 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 12:06:38,834 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9081, 3.5985, 3.5371, 3.5818, 3.6456, 3.4266, 3.7462, 3.3529], + device='cuda:1'), covar=tensor([0.0106, 0.0183, 0.0137, 0.0159, 0.0350, 0.0111, 0.0143, 0.0189], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0107, 0.0108, 0.0092, 0.0184, 0.0114, 0.0110, 0.0119], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 12:06:40,309 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4153, 2.4880, 2.7192, 2.5919, 2.6745, 2.5995, 2.4218, 1.9622], + device='cuda:1'), covar=tensor([0.0396, 0.0470, 0.0295, 0.0165, 0.0342, 0.0386, 0.0325, 0.0331], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0041, 0.0042, 0.0041, 0.0038, 0.0038, 0.0044, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 12:06:47,485 INFO [train.py:935] (1/2) Epoch 46, validation: loss=0.1654, simple_loss=0.2576, pruned_loss=0.03655, over 1622729.00 frames. +2023-03-21 12:06:47,485 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 12:06:49,034 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=127084.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:06:53,544 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 12:07:00,270 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127106.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:07:04,712 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 12:07:05,852 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2596, 2.0955, 2.4122, 2.1455, 2.4085, 2.2006, 2.2090, 1.6364], + device='cuda:1'), covar=tensor([0.0433, 0.0445, 0.0267, 0.0307, 0.0333, 0.0543, 0.0285, 0.0376], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0041, 0.0041, 0.0040, 0.0038, 0.0038, 0.0044, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 12:07:08,363 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3819, 2.5449, 2.3424, 2.4783, 2.5448, 2.2111, 2.5525, 2.3021], + device='cuda:1'), covar=tensor([0.0616, 0.0779, 0.0874, 0.0895, 0.0537, 0.0744, 0.0509, 0.0947], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0061, 0.0070, 0.0062, 0.0058, 0.0065, 0.0059, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:07:11,237 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 12:07:12,657 INFO [train.py:901] (1/2) Epoch 46, batch 50, loss[loss=0.127, simple_loss=0.2141, pruned_loss=0.02, over 7229.00 frames. ], tot_loss[loss=0.1276, simple_loss=0.2096, pruned_loss=0.02281, over 325378.59 frames. ], batch size: 93, lr: 3.60e-03, grad_scale: 8.0 +2023-03-21 12:07:13,639 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 12:07:16,199 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 12:07:17,307 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5897, 3.8346, 3.6195, 3.7441, 3.4273, 3.7867, 4.0506, 4.0478], + device='cuda:1'), covar=tensor([0.0264, 0.0155, 0.0227, 0.0175, 0.0378, 0.0310, 0.0207, 0.0184], + device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0129, 0.0122, 0.0128, 0.0117, 0.0103, 0.0099, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 12:07:23,302 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 +2023-03-21 12:07:24,995 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=127154.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:07:29,002 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.681e+02 1.912e+02 2.297e+02 4.386e+02, threshold=3.823e+02, percent-clipped=1.0 +2023-03-21 12:07:34,612 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 12:07:34,625 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 12:07:34,677 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127172.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:07:39,044 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-21 12:07:39,140 INFO [train.py:901] (1/2) Epoch 46, batch 100, loss[loss=0.1227, simple_loss=0.2038, pruned_loss=0.02082, over 7275.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.2089, pruned_loss=0.02325, over 573218.41 frames. ], batch size: 66, lr: 3.60e-03, grad_scale: 8.0 +2023-03-21 12:07:54,777 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9652, 3.0209, 2.1975, 3.2825, 2.4628, 2.9459, 1.5744, 2.3564], + device='cuda:1'), covar=tensor([0.0658, 0.1199, 0.3207, 0.0874, 0.0673, 0.0725, 0.4368, 0.2030], + device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0256, 0.0274, 0.0267, 0.0265, 0.0262, 0.0227, 0.0255], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:08:04,559 INFO [train.py:901] (1/2) Epoch 46, batch 150, loss[loss=0.1303, simple_loss=0.218, pruned_loss=0.02128, over 6756.00 frames. ], tot_loss[loss=0.128, simple_loss=0.2096, pruned_loss=0.02319, over 765727.13 frames. ], batch size: 106, lr: 3.60e-03, grad_scale: 8.0 +2023-03-21 12:08:08,908 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6834, 3.0189, 2.7437, 2.6869, 2.8226, 2.5502, 2.7671, 2.7039], + device='cuda:1'), covar=tensor([0.0962, 0.0748, 0.0724, 0.1163, 0.0792, 0.0771, 0.0954, 0.0943], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0061, 0.0070, 0.0062, 0.0058, 0.0065, 0.0059, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:08:14,391 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127249.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:08:21,345 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.127e+02 1.712e+02 2.119e+02 2.555e+02 4.679e+02, threshold=4.238e+02, percent-clipped=2.0 +2023-03-21 12:08:30,965 INFO [train.py:901] (1/2) Epoch 46, batch 200, loss[loss=0.1245, simple_loss=0.2088, pruned_loss=0.0201, over 7252.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.209, pruned_loss=0.0232, over 914406.24 frames. ], batch size: 55, lr: 3.60e-03, grad_scale: 8.0 +2023-03-21 12:08:36,567 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 12:08:39,722 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6407, 4.3264, 4.0765, 4.1213, 3.8801, 2.6650, 2.4224, 4.6248], + device='cuda:1'), covar=tensor([0.0045, 0.0087, 0.0093, 0.0069, 0.0105, 0.0529, 0.0531, 0.0051], + device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0096, 0.0118, 0.0101, 0.0135, 0.0138, 0.0130, 0.0108], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 12:08:41,666 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 12:08:45,941 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127310.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:08:47,982 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 12:08:55,729 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1739, 2.6148, 1.9587, 3.0405, 2.8845, 2.8868, 2.4858, 2.7069], + device='cuda:1'), covar=tensor([0.2333, 0.1173, 0.4409, 0.0665, 0.0333, 0.0262, 0.0379, 0.0461], + device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0229, 0.0243, 0.0257, 0.0201, 0.0204, 0.0219, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:08:56,981 INFO [train.py:901] (1/2) Epoch 46, batch 250, loss[loss=0.1487, simple_loss=0.2191, pruned_loss=0.03919, over 7257.00 frames. ], tot_loss[loss=0.1273, simple_loss=0.2087, pruned_loss=0.02295, over 1031939.27 frames. ], batch size: 52, lr: 3.60e-03, grad_scale: 8.0 +2023-03-21 12:08:59,286 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.57 vs. limit=5.0 +2023-03-21 12:09:00,522 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 12:09:13,223 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.252e+02 1.735e+02 1.962e+02 2.246e+02 8.468e+02, threshold=3.924e+02, percent-clipped=3.0 +2023-03-21 12:09:21,355 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 12:09:22,820 INFO [train.py:901] (1/2) Epoch 46, batch 300, loss[loss=0.1193, simple_loss=0.1931, pruned_loss=0.02271, over 7298.00 frames. ], tot_loss[loss=0.1271, simple_loss=0.2084, pruned_loss=0.02285, over 1123603.97 frames. ], batch size: 42, lr: 3.60e-03, grad_scale: 8.0 +2023-03-21 12:09:30,978 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 12:09:47,400 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127427.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 12:09:49,293 INFO [train.py:901] (1/2) Epoch 46, batch 350, loss[loss=0.1025, simple_loss=0.1678, pruned_loss=0.01858, over 6368.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2077, pruned_loss=0.02285, over 1193716.04 frames. ], batch size: 27, lr: 3.60e-03, grad_scale: 8.0 +2023-03-21 12:10:05,546 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.180e+02 1.753e+02 1.971e+02 2.330e+02 3.656e+02, threshold=3.942e+02, percent-clipped=0.0 +2023-03-21 12:10:07,613 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 12:10:10,656 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127472.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:10:12,827 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4801, 3.5006, 2.4735, 3.2194, 2.4914, 2.0431, 1.6138, 3.5296], + device='cuda:1'), covar=tensor([0.0089, 0.0094, 0.0353, 0.0137, 0.0376, 0.0908, 0.0929, 0.0097], + device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0096, 0.0118, 0.0101, 0.0136, 0.0138, 0.0130, 0.0108], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 12:10:15,186 INFO [train.py:901] (1/2) Epoch 46, batch 400, loss[loss=0.1443, simple_loss=0.2236, pruned_loss=0.03249, over 7285.00 frames. ], tot_loss[loss=0.126, simple_loss=0.2073, pruned_loss=0.02228, over 1249451.77 frames. ], batch size: 57, lr: 3.60e-03, grad_scale: 8.0 +2023-03-21 12:10:18,862 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127488.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 12:10:35,742 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=127520.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:10:41,030 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127529.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:10:41,886 INFO [train.py:901] (1/2) Epoch 46, batch 450, loss[loss=0.1184, simple_loss=0.1916, pruned_loss=0.02261, over 7228.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.207, pruned_loss=0.02222, over 1291409.17 frames. ], batch size: 45, lr: 3.60e-03, grad_scale: 8.0 +2023-03-21 12:10:51,107 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 12:10:51,121 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 12:10:57,665 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.232e+02 1.678e+02 2.040e+02 2.414e+02 3.760e+02, threshold=4.080e+02, percent-clipped=0.0 +2023-03-21 12:11:07,473 INFO [train.py:901] (1/2) Epoch 46, batch 500, loss[loss=0.1221, simple_loss=0.2038, pruned_loss=0.02023, over 7350.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.2074, pruned_loss=0.02225, over 1323012.32 frames. ], batch size: 73, lr: 3.60e-03, grad_scale: 8.0 +2023-03-21 12:11:07,656 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1640, 2.4843, 2.0132, 2.9413, 2.7743, 2.7822, 2.5492, 2.6174], + device='cuda:1'), covar=tensor([0.2387, 0.1199, 0.3959, 0.0551, 0.0357, 0.0328, 0.0421, 0.0462], + device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0228, 0.0241, 0.0255, 0.0200, 0.0203, 0.0218, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:11:10,274 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9741, 3.2006, 2.2246, 3.4946, 2.5374, 3.1122, 1.5571, 2.3852], + device='cuda:1'), covar=tensor([0.0535, 0.0818, 0.3105, 0.0693, 0.0512, 0.0874, 0.4413, 0.2055], + device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0255, 0.0274, 0.0265, 0.0264, 0.0260, 0.0227, 0.0254], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:11:12,729 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127590.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:11:20,646 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127605.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:11:23,249 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 12:11:24,787 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 12:11:25,272 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 12:11:27,310 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 12:11:32,364 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 12:11:34,343 INFO [train.py:901] (1/2) Epoch 46, batch 550, loss[loss=0.1116, simple_loss=0.1924, pruned_loss=0.01544, over 7161.00 frames. ], tot_loss[loss=0.1263, simple_loss=0.2078, pruned_loss=0.02236, over 1349428.72 frames. ], batch size: 41, lr: 3.60e-03, grad_scale: 8.0 +2023-03-21 12:11:42,521 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0533, 4.5142, 4.3416, 4.9707, 4.8038, 4.8876, 4.4939, 4.5805], + device='cuda:1'), covar=tensor([0.0829, 0.2514, 0.2312, 0.1077, 0.0954, 0.1156, 0.0809, 0.1101], + device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0403, 0.0305, 0.0323, 0.0240, 0.0377, 0.0238, 0.0287], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 12:11:43,506 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 12:11:50,415 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127662.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:11:50,748 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.225e+02 1.755e+02 2.100e+02 2.513e+02 5.423e+02, threshold=4.201e+02, percent-clipped=2.0 +2023-03-21 12:11:50,811 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 12:11:54,907 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 12:12:00,437 INFO [train.py:901] (1/2) Epoch 46, batch 600, loss[loss=0.1262, simple_loss=0.2077, pruned_loss=0.02235, over 7290.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2081, pruned_loss=0.0226, over 1370095.96 frames. ], batch size: 66, lr: 3.60e-03, grad_scale: 8.0 +2023-03-21 12:12:01,478 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 12:12:02,074 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6324, 1.4683, 1.7489, 1.9967, 1.8355, 2.1247, 1.4398, 2.0558], + device='cuda:1'), covar=tensor([0.2058, 0.3838, 0.1288, 0.1168, 0.1144, 0.1015, 0.2374, 0.1982], + device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0082, 0.0076, 0.0069, 0.0069, 0.0067, 0.0108, 0.0070], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 12:12:04,494 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5859, 4.2840, 4.1850, 4.2766, 4.2102, 4.2592, 4.5023, 4.0279], + device='cuda:1'), covar=tensor([0.0141, 0.0145, 0.0128, 0.0152, 0.0426, 0.0113, 0.0140, 0.0169], + device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0106, 0.0107, 0.0092, 0.0182, 0.0113, 0.0110, 0.0118], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 12:12:17,940 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 12:12:22,110 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127723.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:12:26,000 INFO [train.py:901] (1/2) Epoch 46, batch 650, loss[loss=0.1302, simple_loss=0.2156, pruned_loss=0.02246, over 7361.00 frames. ], tot_loss[loss=0.1265, simple_loss=0.2078, pruned_loss=0.02261, over 1385897.97 frames. ], batch size: 73, lr: 3.59e-03, grad_scale: 8.0 +2023-03-21 12:12:26,534 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 12:12:42,527 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.739e+02 1.949e+02 2.309e+02 6.874e+02, threshold=3.898e+02, percent-clipped=2.0 +2023-03-21 12:12:43,591 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 12:12:51,276 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127779.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:12:52,161 INFO [train.py:901] (1/2) Epoch 46, batch 700, loss[loss=0.1186, simple_loss=0.2081, pruned_loss=0.01451, over 7313.00 frames. ], tot_loss[loss=0.1266, simple_loss=0.2079, pruned_loss=0.02266, over 1397845.92 frames. ], batch size: 83, lr: 3.59e-03, grad_scale: 8.0 +2023-03-21 12:12:52,699 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 12:12:53,289 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127783.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 12:13:12,812 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8619, 2.9518, 3.7565, 3.7591, 3.8211, 3.9039, 3.9515, 3.7655], + device='cuda:1'), covar=tensor([0.0038, 0.0153, 0.0038, 0.0034, 0.0036, 0.0031, 0.0047, 0.0052], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0074, 0.0061, 0.0060, 0.0057, 0.0062, 0.0050, 0.0082], + device='cuda:1'), out_proj_covar=tensor([8.5244e-05, 1.4635e-04, 1.0655e-04, 1.0094e-04, 9.4742e-05, 1.0649e-04, + 9.3936e-05, 1.4858e-04], device='cuda:1') +2023-03-21 12:13:14,888 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7243, 2.8853, 2.5835, 2.9020, 2.7874, 2.4489, 2.8478, 2.7937], + device='cuda:1'), covar=tensor([0.0689, 0.0818, 0.0974, 0.0857, 0.1228, 0.0709, 0.0853, 0.0657], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0061, 0.0070, 0.0061, 0.0058, 0.0065, 0.0059, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:13:16,755 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 12:13:17,237 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 12:13:17,724 INFO [train.py:901] (1/2) Epoch 46, batch 750, loss[loss=0.0999, simple_loss=0.1703, pruned_loss=0.01476, over 6040.00 frames. ], tot_loss[loss=0.1269, simple_loss=0.2084, pruned_loss=0.02272, over 1407499.77 frames. ], batch size: 26, lr: 3.59e-03, grad_scale: 8.0 +2023-03-21 12:13:23,090 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127840.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:13:28,187 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127850.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:13:30,471 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 +2023-03-21 12:13:31,614 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 12:13:34,691 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.337e+02 1.689e+02 1.958e+02 2.466e+02 6.234e+02, threshold=3.916e+02, percent-clipped=1.0 +2023-03-21 12:13:36,877 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 12:13:39,921 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127872.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:13:42,843 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 12:13:43,832 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 12:13:44,337 INFO [train.py:901] (1/2) Epoch 46, batch 800, loss[loss=0.1274, simple_loss=0.2075, pruned_loss=0.02366, over 7327.00 frames. ], tot_loss[loss=0.1276, simple_loss=0.2091, pruned_loss=0.023, over 1417234.31 frames. ], batch size: 61, lr: 3.59e-03, grad_scale: 8.0 +2023-03-21 12:13:46,483 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127885.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:13:50,059 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4822, 2.6204, 3.4105, 3.4088, 3.5080, 3.5642, 3.3810, 3.4289], + device='cuda:1'), covar=tensor([0.0034, 0.0178, 0.0038, 0.0039, 0.0036, 0.0031, 0.0085, 0.0058], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0074, 0.0061, 0.0060, 0.0057, 0.0063, 0.0050, 0.0082], + device='cuda:1'), out_proj_covar=tensor([8.5465e-05, 1.4671e-04, 1.0640e-04, 1.0113e-04, 9.4912e-05, 1.0677e-04, + 9.4352e-05, 1.4904e-04], device='cuda:1') +2023-03-21 12:13:54,462 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 12:13:56,524 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127905.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:13:59,523 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127911.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:14:09,823 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 +2023-03-21 12:14:09,941 INFO [train.py:901] (1/2) Epoch 46, batch 850, loss[loss=0.1198, simple_loss=0.1957, pruned_loss=0.02191, over 7231.00 frames. ], tot_loss[loss=0.1275, simple_loss=0.2091, pruned_loss=0.02301, over 1423301.19 frames. ], batch size: 45, lr: 3.59e-03, grad_scale: 8.0 +2023-03-21 12:14:11,134 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127933.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:14:13,058 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 12:14:13,546 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 12:14:18,131 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 12:14:21,932 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=127953.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:14:22,883 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 12:14:25,257 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 +2023-03-21 12:14:26,865 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.179e+02 1.746e+02 1.944e+02 2.264e+02 3.850e+02, threshold=3.887e+02, percent-clipped=0.0 +2023-03-21 12:14:35,893 INFO [train.py:901] (1/2) Epoch 46, batch 900, loss[loss=0.1124, simple_loss=0.1946, pruned_loss=0.01514, over 7223.00 frames. ], tot_loss[loss=0.1276, simple_loss=0.2089, pruned_loss=0.02311, over 1426799.55 frames. ], batch size: 45, lr: 3.59e-03, grad_scale: 8.0 +2023-03-21 12:14:39,642 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1543, 2.7956, 3.2128, 3.1390, 3.1986, 2.9490, 2.5461, 3.0757], + device='cuda:1'), covar=tensor([0.0938, 0.0799, 0.1062, 0.1035, 0.0876, 0.1505, 0.1854, 0.1164], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0071, 0.0053, 0.0052, 0.0052, 0.0051, 0.0070, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 12:14:59,353 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128018.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:15:03,393 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 12:15:06,546 INFO [train.py:901] (1/2) Epoch 46, batch 950, loss[loss=0.1164, simple_loss=0.2029, pruned_loss=0.01496, over 7250.00 frames. ], tot_loss[loss=0.1272, simple_loss=0.2089, pruned_loss=0.02281, over 1432176.97 frames. ], batch size: 55, lr: 3.59e-03, grad_scale: 8.0 +2023-03-21 12:15:22,629 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.200e+02 1.644e+02 1.867e+02 2.236e+02 5.162e+02, threshold=3.734e+02, percent-clipped=3.0 +2023-03-21 12:15:26,713 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 12:15:31,753 INFO [train.py:901] (1/2) Epoch 46, batch 1000, loss[loss=0.1276, simple_loss=0.2065, pruned_loss=0.02431, over 7265.00 frames. ], tot_loss[loss=0.1272, simple_loss=0.2086, pruned_loss=0.0229, over 1435201.52 frames. ], batch size: 70, lr: 3.59e-03, grad_scale: 8.0 +2023-03-21 12:15:32,819 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128083.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 12:15:42,382 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7141, 1.4727, 2.1313, 2.1925, 2.0817, 2.2222, 1.8112, 2.3487], + device='cuda:1'), covar=tensor([0.2864, 0.2952, 0.1561, 0.0879, 0.1328, 0.1520, 0.1490, 0.1477], + device='cuda:1'), in_proj_covar=tensor([0.0084, 0.0082, 0.0076, 0.0069, 0.0069, 0.0067, 0.0108, 0.0070], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 12:15:45,462 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0777, 1.9987, 2.1389, 2.0544, 2.2013, 2.0726, 1.9717, 1.5917], + device='cuda:1'), covar=tensor([0.0516, 0.0458, 0.0452, 0.0363, 0.0476, 0.0490, 0.0547, 0.0386], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0041, 0.0042, 0.0041, 0.0038, 0.0039, 0.0045, 0.0044], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 12:15:47,832 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 12:15:58,210 INFO [train.py:901] (1/2) Epoch 46, batch 1050, loss[loss=0.1389, simple_loss=0.2198, pruned_loss=0.02897, over 7319.00 frames. ], tot_loss[loss=0.1273, simple_loss=0.2091, pruned_loss=0.02277, over 1437713.68 frames. ], batch size: 75, lr: 3.59e-03, grad_scale: 8.0 +2023-03-21 12:15:58,274 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=128131.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 12:16:00,280 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128135.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:16:08,848 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 12:16:12,364 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 12:16:14,377 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.236e+02 1.762e+02 2.099e+02 2.466e+02 4.105e+02, threshold=4.197e+02, percent-clipped=1.0 +2023-03-21 12:16:24,064 INFO [train.py:901] (1/2) Epoch 46, batch 1100, loss[loss=0.1284, simple_loss=0.2111, pruned_loss=0.02279, over 7287.00 frames. ], tot_loss[loss=0.1274, simple_loss=0.2092, pruned_loss=0.02278, over 1439905.90 frames. ], batch size: 66, lr: 3.59e-03, grad_scale: 8.0 +2023-03-21 12:16:26,219 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128185.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:16:37,755 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128206.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:16:42,773 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 12:16:43,230 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 12:16:48,863 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128228.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:16:50,334 INFO [train.py:901] (1/2) Epoch 46, batch 1150, loss[loss=0.1154, simple_loss=0.1842, pruned_loss=0.02337, over 6971.00 frames. ], tot_loss[loss=0.1278, simple_loss=0.2096, pruned_loss=0.02299, over 1441880.28 frames. ], batch size: 35, lr: 3.59e-03, grad_scale: 8.0 +2023-03-21 12:16:51,387 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=128233.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:16:56,334 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 12:16:56,846 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 12:17:06,364 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.776e+02 2.012e+02 2.405e+02 4.369e+02, threshold=4.024e+02, percent-clipped=1.0 +2023-03-21 12:17:16,094 INFO [train.py:901] (1/2) Epoch 46, batch 1200, loss[loss=0.1567, simple_loss=0.2265, pruned_loss=0.04343, over 7252.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.2093, pruned_loss=0.02308, over 1440875.75 frames. ], batch size: 55, lr: 3.59e-03, grad_scale: 8.0 +2023-03-21 12:17:31,506 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 12:17:35,665 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128318.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:17:36,634 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128320.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:17:42,002 INFO [train.py:901] (1/2) Epoch 46, batch 1250, loss[loss=0.1143, simple_loss=0.2007, pruned_loss=0.01389, over 7351.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.2091, pruned_loss=0.02318, over 1442337.69 frames. ], batch size: 73, lr: 3.59e-03, grad_scale: 8.0 +2023-03-21 12:17:43,856 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-03-21 12:17:49,599 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.81 vs. limit=5.0 +2023-03-21 12:17:55,477 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 12:17:59,031 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.390e+02 1.796e+02 1.977e+02 2.380e+02 4.062e+02, threshold=3.955e+02, percent-clipped=1.0 +2023-03-21 12:18:00,575 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 12:18:00,620 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=128366.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:18:01,590 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 12:18:08,157 INFO [train.py:901] (1/2) Epoch 46, batch 1300, loss[loss=0.1199, simple_loss=0.2097, pruned_loss=0.01504, over 7279.00 frames. ], tot_loss[loss=0.1276, simple_loss=0.2089, pruned_loss=0.02311, over 1440683.42 frames. ], batch size: 77, lr: 3.59e-03, grad_scale: 8.0 +2023-03-21 12:18:08,299 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128381.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:18:25,361 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128413.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:18:25,759 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 12:18:28,339 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 12:18:31,832 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 12:18:34,382 INFO [train.py:901] (1/2) Epoch 46, batch 1350, loss[loss=0.1234, simple_loss=0.2093, pruned_loss=0.01879, over 7315.00 frames. ], tot_loss[loss=0.1272, simple_loss=0.2086, pruned_loss=0.02287, over 1440461.74 frames. ], batch size: 80, lr: 3.59e-03, grad_scale: 8.0 +2023-03-21 12:18:36,593 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128435.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:18:38,181 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1708, 3.0525, 3.3584, 2.9562, 3.3238, 3.2702, 2.7775, 3.0879], + device='cuda:1'), covar=tensor([0.1252, 0.0812, 0.0971, 0.2231, 0.0992, 0.0922, 0.1693, 0.1544], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0071, 0.0053, 0.0052, 0.0053, 0.0051, 0.0070, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 12:18:40,569 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 12:18:51,194 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.207e+02 1.598e+02 1.958e+02 2.374e+02 5.805e+02, threshold=3.916e+02, percent-clipped=3.0 +2023-03-21 12:18:57,560 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128474.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:19:00,955 INFO [train.py:901] (1/2) Epoch 46, batch 1400, loss[loss=0.1398, simple_loss=0.2236, pruned_loss=0.02799, over 7221.00 frames. ], tot_loss[loss=0.1276, simple_loss=0.2092, pruned_loss=0.02297, over 1442793.03 frames. ], batch size: 93, lr: 3.58e-03, grad_scale: 8.0 +2023-03-21 12:19:02,041 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=128483.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:19:13,828 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128506.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:19:14,242 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 12:19:24,947 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128528.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:19:26,278 INFO [train.py:901] (1/2) Epoch 46, batch 1450, loss[loss=0.1332, simple_loss=0.2192, pruned_loss=0.02359, over 7303.00 frames. ], tot_loss[loss=0.1274, simple_loss=0.209, pruned_loss=0.02289, over 1441788.78 frames. ], batch size: 80, lr: 3.58e-03, grad_scale: 8.0 +2023-03-21 12:19:38,595 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 12:19:39,165 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=128554.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:19:43,534 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.251e+02 1.693e+02 1.942e+02 2.319e+02 7.127e+02, threshold=3.884e+02, percent-clipped=1.0 +2023-03-21 12:19:43,728 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3372, 3.5367, 2.4759, 3.9605, 3.0965, 3.2794, 1.7822, 2.7374], + device='cuda:1'), covar=tensor([0.0453, 0.0878, 0.2923, 0.0535, 0.0508, 0.0656, 0.4212, 0.1927], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0254, 0.0274, 0.0266, 0.0263, 0.0261, 0.0226, 0.0254], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:19:50,121 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=128576.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:19:52,615 INFO [train.py:901] (1/2) Epoch 46, batch 1500, loss[loss=0.1304, simple_loss=0.2148, pruned_loss=0.02303, over 7247.00 frames. ], tot_loss[loss=0.1276, simple_loss=0.2094, pruned_loss=0.02295, over 1444097.52 frames. ], batch size: 55, lr: 3.58e-03, grad_scale: 8.0 +2023-03-21 12:19:54,700 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 12:20:15,480 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 +2023-03-21 12:20:18,681 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 12:20:19,184 INFO [train.py:901] (1/2) Epoch 46, batch 1550, loss[loss=0.1008, simple_loss=0.1813, pruned_loss=0.01017, over 7174.00 frames. ], tot_loss[loss=0.1274, simple_loss=0.2089, pruned_loss=0.02294, over 1442098.90 frames. ], batch size: 39, lr: 3.58e-03, grad_scale: 8.0 +2023-03-21 12:20:35,803 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 1.729e+02 1.986e+02 2.320e+02 3.643e+02, threshold=3.971e+02, percent-clipped=0.0 +2023-03-21 12:20:42,474 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128676.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:20:44,916 INFO [train.py:901] (1/2) Epoch 46, batch 1600, loss[loss=0.1273, simple_loss=0.2135, pruned_loss=0.02057, over 7276.00 frames. ], tot_loss[loss=0.1275, simple_loss=0.209, pruned_loss=0.02305, over 1443298.96 frames. ], batch size: 47, lr: 3.58e-03, grad_scale: 8.0 +2023-03-21 12:20:49,946 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 12:20:50,955 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 12:20:53,445 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 12:20:57,627 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7632, 2.9465, 2.6359, 2.8625, 2.8392, 2.5325, 2.8859, 2.7993], + device='cuda:1'), covar=tensor([0.0718, 0.0782, 0.1077, 0.1070, 0.1282, 0.1028, 0.0751, 0.1124], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0061, 0.0070, 0.0062, 0.0058, 0.0065, 0.0059, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:21:03,820 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 12:21:08,592 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 12:21:08,710 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2440, 2.2480, 2.5933, 2.2656, 2.4427, 2.4470, 2.3295, 1.8886], + device='cuda:1'), covar=tensor([0.0540, 0.0564, 0.0378, 0.0388, 0.0505, 0.0389, 0.0445, 0.0382], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0042, 0.0042, 0.0041, 0.0039, 0.0040, 0.0046, 0.0045], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 12:21:11,661 INFO [train.py:901] (1/2) Epoch 46, batch 1650, loss[loss=0.1188, simple_loss=0.2059, pruned_loss=0.01589, over 7274.00 frames. ], tot_loss[loss=0.1277, simple_loss=0.2091, pruned_loss=0.02317, over 1443009.84 frames. ], batch size: 66, lr: 3.58e-03, grad_scale: 8.0 +2023-03-21 12:21:17,757 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 12:21:27,885 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.187e+02 1.643e+02 1.877e+02 2.168e+02 3.378e+02, threshold=3.754e+02, percent-clipped=0.0 +2023-03-21 12:21:31,009 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128769.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:21:33,449 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 12:21:35,517 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8396, 4.0195, 3.7397, 3.9502, 3.5826, 3.9585, 4.2625, 4.2628], + device='cuda:1'), covar=tensor([0.0239, 0.0155, 0.0260, 0.0158, 0.0428, 0.0285, 0.0228, 0.0207], + device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0129, 0.0123, 0.0128, 0.0117, 0.0103, 0.0100, 0.0105], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 12:21:36,910 INFO [train.py:901] (1/2) Epoch 46, batch 1700, loss[loss=0.1298, simple_loss=0.2114, pruned_loss=0.02408, over 7272.00 frames. ], tot_loss[loss=0.1278, simple_loss=0.2092, pruned_loss=0.02324, over 1442734.22 frames. ], batch size: 89, lr: 3.58e-03, grad_scale: 8.0 +2023-03-21 12:21:37,053 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128781.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:21:37,956 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 12:21:50,008 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 12:22:04,343 INFO [train.py:901] (1/2) Epoch 46, batch 1750, loss[loss=0.1205, simple_loss=0.2084, pruned_loss=0.01625, over 7242.00 frames. ], tot_loss[loss=0.1273, simple_loss=0.2084, pruned_loss=0.02307, over 1439678.85 frames. ], batch size: 89, lr: 3.58e-03, grad_scale: 8.0 +2023-03-21 12:22:10,010 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128842.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:22:15,979 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 12:22:16,993 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 12:22:20,528 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.769e+02 2.041e+02 2.508e+02 4.392e+02, threshold=4.083e+02, percent-clipped=1.0 +2023-03-21 12:22:21,688 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9533, 2.6075, 3.2848, 2.9092, 3.0952, 2.8679, 2.5245, 3.0528], + device='cuda:1'), covar=tensor([0.1381, 0.0716, 0.0735, 0.1119, 0.0703, 0.0988, 0.1676, 0.1056], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0070, 0.0053, 0.0052, 0.0052, 0.0051, 0.0069, 0.0052], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 12:22:29,559 INFO [train.py:901] (1/2) Epoch 46, batch 1800, loss[loss=0.1285, simple_loss=0.2014, pruned_loss=0.02781, over 7230.00 frames. ], tot_loss[loss=0.1274, simple_loss=0.2088, pruned_loss=0.02305, over 1442387.83 frames. ], batch size: 50, lr: 3.58e-03, grad_scale: 8.0 +2023-03-21 12:22:34,999 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128890.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 12:22:38,399 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 12:22:51,863 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 12:22:56,420 INFO [train.py:901] (1/2) Epoch 46, batch 1850, loss[loss=0.128, simple_loss=0.213, pruned_loss=0.02149, over 7252.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2081, pruned_loss=0.02239, over 1443159.84 frames. ], batch size: 55, lr: 3.58e-03, grad_scale: 8.0 +2023-03-21 12:23:00,365 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 +2023-03-21 12:23:02,476 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 12:23:06,760 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128951.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 12:23:12,627 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 1.665e+02 1.878e+02 2.204e+02 4.262e+02, threshold=3.756e+02, percent-clipped=1.0 +2023-03-21 12:23:19,326 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 12:23:19,942 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128976.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:23:22,339 INFO [train.py:901] (1/2) Epoch 46, batch 1900, loss[loss=0.1325, simple_loss=0.2213, pruned_loss=0.02186, over 7276.00 frames. ], tot_loss[loss=0.1269, simple_loss=0.2082, pruned_loss=0.02277, over 1442064.86 frames. ], batch size: 64, lr: 3.58e-03, grad_scale: 8.0 +2023-03-21 12:23:29,755 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5822, 1.8012, 1.5369, 1.6506, 1.8267, 1.7655, 1.7050, 1.4478], + device='cuda:1'), covar=tensor([0.0197, 0.0163, 0.0297, 0.0280, 0.0132, 0.0172, 0.0169, 0.0211], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0040, 0.0038, 0.0040, 0.0038, 0.0037, 0.0039, 0.0048], + device='cuda:1'), out_proj_covar=tensor([4.5611e-05, 4.4775e-05, 4.2750e-05, 4.4164e-05, 4.2354e-05, 4.0503e-05, + 4.3323e-05, 5.2563e-05], device='cuda:1') +2023-03-21 12:23:44,392 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 12:23:45,433 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=129024.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:23:49,021 INFO [train.py:901] (1/2) Epoch 46, batch 1950, loss[loss=0.156, simple_loss=0.2302, pruned_loss=0.04091, over 7281.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2083, pruned_loss=0.02262, over 1442272.50 frames. ], batch size: 57, lr: 3.58e-03, grad_scale: 8.0 +2023-03-21 12:23:56,704 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 12:24:01,401 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 12:24:02,402 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 12:24:05,974 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.250e+02 1.809e+02 2.110e+02 2.690e+02 9.152e+02, threshold=4.221e+02, percent-clipped=6.0 +2023-03-21 12:24:09,114 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129069.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:24:15,731 INFO [train.py:901] (1/2) Epoch 46, batch 2000, loss[loss=0.1356, simple_loss=0.2172, pruned_loss=0.02698, over 7248.00 frames. ], tot_loss[loss=0.1269, simple_loss=0.2085, pruned_loss=0.02261, over 1443363.38 frames. ], batch size: 55, lr: 3.58e-03, grad_scale: 8.0 +2023-03-21 12:24:19,907 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 12:24:30,973 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 12:24:34,070 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=129117.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:24:38,636 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 12:24:41,086 INFO [train.py:901] (1/2) Epoch 46, batch 2050, loss[loss=0.1322, simple_loss=0.2089, pruned_loss=0.02778, over 7360.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2085, pruned_loss=0.02245, over 1444468.93 frames. ], batch size: 51, lr: 3.58e-03, grad_scale: 8.0 +2023-03-21 12:24:44,162 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129137.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:24:57,814 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.071e+02 1.744e+02 1.999e+02 2.402e+02 4.134e+02, threshold=3.998e+02, percent-clipped=0.0 +2023-03-21 12:25:07,485 INFO [train.py:901] (1/2) Epoch 46, batch 2100, loss[loss=0.1242, simple_loss=0.2083, pruned_loss=0.0201, over 7290.00 frames. ], tot_loss[loss=0.1265, simple_loss=0.2084, pruned_loss=0.0223, over 1444247.14 frames. ], batch size: 86, lr: 3.57e-03, grad_scale: 8.0 +2023-03-21 12:25:12,548 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 12:25:15,580 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 12:25:15,885 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-03-21 12:25:33,578 INFO [train.py:901] (1/2) Epoch 46, batch 2150, loss[loss=0.1184, simple_loss=0.2072, pruned_loss=0.01478, over 7361.00 frames. ], tot_loss[loss=0.1261, simple_loss=0.208, pruned_loss=0.0221, over 1443686.42 frames. ], batch size: 63, lr: 3.57e-03, grad_scale: 8.0 +2023-03-21 12:25:42,513 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129246.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 12:25:50,976 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.299e+02 1.747e+02 2.060e+02 2.463e+02 4.425e+02, threshold=4.120e+02, percent-clipped=1.0 +2023-03-21 12:26:00,050 INFO [train.py:901] (1/2) Epoch 46, batch 2200, loss[loss=0.1396, simple_loss=0.2157, pruned_loss=0.03174, over 7260.00 frames. ], tot_loss[loss=0.127, simple_loss=0.2088, pruned_loss=0.02262, over 1445321.97 frames. ], batch size: 52, lr: 3.57e-03, grad_scale: 8.0 +2023-03-21 12:26:02,533 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 12:26:16,795 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9792, 3.6691, 3.6668, 3.6645, 3.6947, 3.5986, 3.8531, 3.4849], + device='cuda:1'), covar=tensor([0.0152, 0.0197, 0.0152, 0.0190, 0.0391, 0.0127, 0.0146, 0.0188], + device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0105, 0.0107, 0.0091, 0.0181, 0.0113, 0.0108, 0.0117], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 12:26:26,029 INFO [train.py:901] (1/2) Epoch 46, batch 2250, loss[loss=0.1415, simple_loss=0.2196, pruned_loss=0.03174, over 7362.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2082, pruned_loss=0.0223, over 1444791.14 frames. ], batch size: 63, lr: 3.57e-03, grad_scale: 8.0 +2023-03-21 12:26:37,746 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 12:26:37,763 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 12:26:42,749 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.311e+02 1.693e+02 1.987e+02 2.200e+02 3.291e+02, threshold=3.973e+02, percent-clipped=0.0 +2023-03-21 12:26:50,355 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 12:26:51,829 INFO [train.py:901] (1/2) Epoch 46, batch 2300, loss[loss=0.08739, simple_loss=0.153, pruned_loss=0.01087, over 5936.00 frames. ], tot_loss[loss=0.1269, simple_loss=0.2085, pruned_loss=0.02266, over 1441961.55 frames. ], batch size: 25, lr: 3.57e-03, grad_scale: 8.0 +2023-03-21 12:27:09,030 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129413.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 12:27:18,754 INFO [train.py:901] (1/2) Epoch 46, batch 2350, loss[loss=0.1207, simple_loss=0.2046, pruned_loss=0.01841, over 7282.00 frames. ], tot_loss[loss=0.1269, simple_loss=0.2082, pruned_loss=0.02277, over 1443971.01 frames. ], batch size: 70, lr: 3.57e-03, grad_scale: 8.0 +2023-03-21 12:27:21,878 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129437.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:27:34,992 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.209e+02 1.731e+02 2.025e+02 2.422e+02 3.708e+02, threshold=4.049e+02, percent-clipped=0.0 +2023-03-21 12:27:37,110 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 12:27:40,769 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129474.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 12:27:43,710 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 12:27:44,210 INFO [train.py:901] (1/2) Epoch 46, batch 2400, loss[loss=0.1021, simple_loss=0.1777, pruned_loss=0.01321, over 7207.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2082, pruned_loss=0.02261, over 1440481.74 frames. ], batch size: 39, lr: 3.57e-03, grad_scale: 8.0 +2023-03-21 12:27:46,357 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=129485.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:27:53,215 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-03-21 12:27:53,957 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 12:27:57,001 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 12:28:06,999 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9422, 2.7730, 3.1219, 3.3129, 2.8336, 2.8285, 3.2753, 2.3178], + device='cuda:1'), covar=tensor([0.0407, 0.0533, 0.0877, 0.0719, 0.0807, 0.1107, 0.0591, 0.2834], + device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0334, 0.0267, 0.0347, 0.0279, 0.0283, 0.0342, 0.0235], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 12:28:10,908 INFO [train.py:901] (1/2) Epoch 46, batch 2450, loss[loss=0.1328, simple_loss=0.2152, pruned_loss=0.02519, over 7345.00 frames. ], tot_loss[loss=0.1265, simple_loss=0.2079, pruned_loss=0.02252, over 1441525.68 frames. ], batch size: 73, lr: 3.57e-03, grad_scale: 8.0 +2023-03-21 12:28:17,977 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-03-21 12:28:18,686 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129546.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 12:28:24,813 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 12:28:27,228 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.108e+02 1.840e+02 2.063e+02 2.420e+02 3.694e+02, threshold=4.127e+02, percent-clipped=0.0 +2023-03-21 12:28:36,372 INFO [train.py:901] (1/2) Epoch 46, batch 2500, loss[loss=0.1448, simple_loss=0.2272, pruned_loss=0.03122, over 7306.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2079, pruned_loss=0.02244, over 1440364.58 frames. ], batch size: 83, lr: 3.57e-03, grad_scale: 8.0 +2023-03-21 12:28:43,658 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=129594.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 12:28:52,155 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 12:28:53,877 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129612.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:29:03,265 INFO [train.py:901] (1/2) Epoch 46, batch 2550, loss[loss=0.1348, simple_loss=0.2135, pruned_loss=0.02807, over 7260.00 frames. ], tot_loss[loss=0.1271, simple_loss=0.2089, pruned_loss=0.02267, over 1442912.24 frames. ], batch size: 64, lr: 3.57e-03, grad_scale: 16.0 +2023-03-21 12:29:05,664 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 +2023-03-21 12:29:16,236 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6105, 3.0173, 2.6741, 2.9958, 2.8723, 2.4909, 2.9429, 2.5945], + device='cuda:1'), covar=tensor([0.0968, 0.0679, 0.1010, 0.0786, 0.1383, 0.0912, 0.0663, 0.1135], + device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0061, 0.0070, 0.0062, 0.0058, 0.0065, 0.0059, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:29:19,657 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.370e+02 1.850e+02 2.116e+02 2.396e+02 5.730e+02, threshold=4.232e+02, percent-clipped=1.0 +2023-03-21 12:29:24,877 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129673.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:29:29,332 INFO [train.py:901] (1/2) Epoch 46, batch 2600, loss[loss=0.131, simple_loss=0.2201, pruned_loss=0.02094, over 7121.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.2076, pruned_loss=0.02213, over 1443539.69 frames. ], batch size: 98, lr: 3.57e-03, grad_scale: 16.0 +2023-03-21 12:29:54,722 INFO [train.py:901] (1/2) Epoch 46, batch 2650, loss[loss=0.1041, simple_loss=0.1854, pruned_loss=0.01141, over 7319.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.208, pruned_loss=0.02246, over 1444296.18 frames. ], batch size: 44, lr: 3.57e-03, grad_scale: 8.0 +2023-03-21 12:30:03,906 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-03-21 12:30:10,985 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.153e+02 1.766e+02 2.133e+02 2.471e+02 4.108e+02, threshold=4.267e+02, percent-clipped=0.0 +2023-03-21 12:30:13,546 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129769.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 12:30:19,403 INFO [train.py:901] (1/2) Epoch 46, batch 2700, loss[loss=0.1212, simple_loss=0.2056, pruned_loss=0.01841, over 7274.00 frames. ], tot_loss[loss=0.127, simple_loss=0.2085, pruned_loss=0.02278, over 1443406.15 frames. ], batch size: 52, lr: 3.57e-03, grad_scale: 8.0 +2023-03-21 12:30:24,491 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129791.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:30:44,368 INFO [train.py:901] (1/2) Epoch 46, batch 2750, loss[loss=0.1258, simple_loss=0.204, pruned_loss=0.0238, over 7264.00 frames. ], tot_loss[loss=0.1273, simple_loss=0.2091, pruned_loss=0.02277, over 1443782.37 frames. ], batch size: 47, lr: 3.57e-03, grad_scale: 8.0 +2023-03-21 12:30:54,947 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129852.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:31:00,577 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.068e+02 1.744e+02 2.096e+02 2.375e+02 4.011e+02, threshold=4.191e+02, percent-clipped=0.0 +2023-03-21 12:31:09,055 INFO [train.py:901] (1/2) Epoch 46, batch 2800, loss[loss=0.1244, simple_loss=0.2173, pruned_loss=0.01578, over 7281.00 frames. ], tot_loss[loss=0.1271, simple_loss=0.2087, pruned_loss=0.0227, over 1443726.66 frames. ], batch size: 77, lr: 3.57e-03, grad_scale: 8.0 +2023-03-21 12:31:31,873 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 12:31:31,924 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 12:31:31,949 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 12:31:31,966 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 12:31:32,144 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 12:31:32,458 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 12:31:32,469 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 12:31:32,564 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 12:31:32,637 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 12:31:32,652 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 12:31:32,673 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 12:31:32,830 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 12:31:32,836 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 12:31:32,843 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 12:31:32,984 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 12:31:32,986 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 12:31:33,043 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 12:31:37,231 INFO [train.py:901] (1/2) Epoch 47, batch 0, loss[loss=0.1344, simple_loss=0.2239, pruned_loss=0.02245, over 7197.00 frames. ], tot_loss[loss=0.1344, simple_loss=0.2239, pruned_loss=0.02245, over 7197.00 frames. ], batch size: 93, lr: 3.53e-03, grad_scale: 8.0 +2023-03-21 12:31:37,232 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 12:31:48,219 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8813, 3.6697, 3.5250, 3.5617, 3.6898, 3.4454, 3.6807, 3.4333], + device='cuda:1'), covar=tensor([0.0114, 0.0152, 0.0126, 0.0162, 0.0282, 0.0113, 0.0155, 0.0143], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0105, 0.0108, 0.0092, 0.0181, 0.0112, 0.0109, 0.0118], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 12:32:02,789 INFO [train.py:935] (1/2) Epoch 47, validation: loss=0.166, simple_loss=0.2594, pruned_loss=0.0363, over 1622729.00 frames. +2023-03-21 12:32:02,790 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 12:32:09,859 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 12:32:20,546 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 12:32:28,824 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 12:32:29,322 INFO [train.py:901] (1/2) Epoch 47, batch 50, loss[loss=0.1096, simple_loss=0.188, pruned_loss=0.01558, over 7146.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2084, pruned_loss=0.02221, over 326532.62 frames. ], batch size: 41, lr: 3.53e-03, grad_scale: 8.0 +2023-03-21 12:32:29,990 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2400, 2.2232, 2.2561, 2.0796, 2.3357, 2.2548, 2.1686, 1.7766], + device='cuda:1'), covar=tensor([0.0467, 0.0459, 0.0471, 0.0442, 0.0617, 0.0508, 0.0366, 0.0421], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0043, 0.0043, 0.0043, 0.0040, 0.0041, 0.0047, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 12:32:30,895 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 12:32:33,927 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.321e+02 1.737e+02 2.014e+02 2.442e+02 5.110e+02, threshold=4.029e+02, percent-clipped=1.0 +2023-03-21 12:32:33,950 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 12:32:36,003 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129968.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:32:49,977 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 12:32:50,441 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 12:32:52,411 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130000.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:32:53,438 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130002.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:32:54,821 INFO [train.py:901] (1/2) Epoch 47, batch 100, loss[loss=0.1339, simple_loss=0.2162, pruned_loss=0.02576, over 7291.00 frames. ], tot_loss[loss=0.1258, simple_loss=0.2076, pruned_loss=0.02198, over 575820.96 frames. ], batch size: 57, lr: 3.53e-03, grad_scale: 8.0 +2023-03-21 12:32:56,643 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-03-21 12:33:08,481 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4671, 2.8007, 2.4911, 2.7742, 2.6691, 2.3549, 2.6563, 2.5430], + device='cuda:1'), covar=tensor([0.0805, 0.0655, 0.0975, 0.0578, 0.0860, 0.0770, 0.1098, 0.0639], + device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0061, 0.0070, 0.0062, 0.0058, 0.0065, 0.0059, 0.0057], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:33:11,149 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-03-21 12:33:21,481 INFO [train.py:901] (1/2) Epoch 47, batch 150, loss[loss=0.1213, simple_loss=0.2051, pruned_loss=0.01882, over 7310.00 frames. ], tot_loss[loss=0.1265, simple_loss=0.2082, pruned_loss=0.02233, over 768666.40 frames. ], batch size: 86, lr: 3.52e-03, grad_scale: 8.0 +2023-03-21 12:33:24,617 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130061.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:33:25,650 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130063.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 12:33:25,989 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.052e+02 1.724e+02 1.922e+02 2.327e+02 4.445e+02, threshold=3.845e+02, percent-clipped=1.0 +2023-03-21 12:33:26,163 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4124, 1.6709, 1.4733, 1.5889, 1.6338, 1.5498, 1.4549, 1.3914], + device='cuda:1'), covar=tensor([0.0259, 0.0154, 0.0196, 0.0151, 0.0130, 0.0205, 0.0190, 0.0159], + device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0041, 0.0038, 0.0040, 0.0038, 0.0037, 0.0039, 0.0048], + device='cuda:1'), out_proj_covar=tensor([4.5507e-05, 4.4877e-05, 4.2477e-05, 4.4053e-05, 4.2294e-05, 4.0665e-05, + 4.3475e-05, 5.2105e-05], device='cuda:1') +2023-03-21 12:33:28,630 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130069.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 12:33:29,606 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.6723, 4.1344, 3.9913, 4.6557, 4.4600, 4.5656, 4.1330, 4.1374], + device='cuda:1'), covar=tensor([0.0924, 0.2853, 0.2759, 0.1103, 0.1010, 0.1312, 0.0902, 0.1316], + device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0406, 0.0304, 0.0322, 0.0235, 0.0380, 0.0235, 0.0286], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 12:33:47,461 INFO [train.py:901] (1/2) Epoch 47, batch 200, loss[loss=0.1392, simple_loss=0.2073, pruned_loss=0.03556, over 7343.00 frames. ], tot_loss[loss=0.1265, simple_loss=0.2079, pruned_loss=0.02255, over 916861.75 frames. ], batch size: 51, lr: 3.52e-03, grad_scale: 8.0 +2023-03-21 12:33:47,849 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 +2023-03-21 12:33:51,878 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 12:33:54,018 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=130117.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 12:33:57,505 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 12:34:03,533 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 12:34:09,153 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130147.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:34:13,044 INFO [train.py:901] (1/2) Epoch 47, batch 250, loss[loss=0.1377, simple_loss=0.219, pruned_loss=0.0282, over 7312.00 frames. ], tot_loss[loss=0.1263, simple_loss=0.2078, pruned_loss=0.02239, over 1032499.47 frames. ], batch size: 59, lr: 3.52e-03, grad_scale: 8.0 +2023-03-21 12:34:16,681 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 12:34:17,635 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.342e+02 1.845e+02 2.146e+02 2.522e+02 4.648e+02, threshold=4.292e+02, percent-clipped=1.0 +2023-03-21 12:34:38,997 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 12:34:39,493 INFO [train.py:901] (1/2) Epoch 47, batch 300, loss[loss=0.1033, simple_loss=0.1792, pruned_loss=0.01374, over 6975.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.2073, pruned_loss=0.02222, over 1124394.96 frames. ], batch size: 35, lr: 3.52e-03, grad_scale: 8.0 +2023-03-21 12:34:49,082 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 12:34:49,339 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-03-21 12:35:00,700 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7116, 2.8982, 3.7827, 3.6226, 3.6770, 3.7863, 3.7621, 3.6917], + device='cuda:1'), covar=tensor([0.0036, 0.0150, 0.0034, 0.0043, 0.0037, 0.0035, 0.0057, 0.0051], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0074, 0.0062, 0.0061, 0.0057, 0.0063, 0.0050, 0.0082], + device='cuda:1'), out_proj_covar=tensor([8.6351e-05, 1.4709e-04, 1.0836e-04, 1.0203e-04, 9.3600e-05, 1.0766e-04, + 9.3052e-05, 1.4860e-04], device='cuda:1') +2023-03-21 12:35:05,233 INFO [train.py:901] (1/2) Epoch 47, batch 350, loss[loss=0.1115, simple_loss=0.1909, pruned_loss=0.016, over 7216.00 frames. ], tot_loss[loss=0.1258, simple_loss=0.2068, pruned_loss=0.02234, over 1193645.81 frames. ], batch size: 45, lr: 3.52e-03, grad_scale: 8.0 +2023-03-21 12:35:09,770 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.604e+02 1.967e+02 2.277e+02 3.840e+02, threshold=3.934e+02, percent-clipped=0.0 +2023-03-21 12:35:11,897 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130268.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:35:16,473 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8272, 2.8102, 2.8524, 3.0252, 2.6281, 2.6681, 3.0621, 2.1305], + device='cuda:1'), covar=tensor([0.0719, 0.0927, 0.1012, 0.0949, 0.0895, 0.1291, 0.0945, 0.2988], + device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0333, 0.0265, 0.0346, 0.0276, 0.0281, 0.0338, 0.0233], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 12:35:23,421 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 12:35:31,569 INFO [train.py:901] (1/2) Epoch 47, batch 400, loss[loss=0.1079, simple_loss=0.1836, pruned_loss=0.01613, over 7175.00 frames. ], tot_loss[loss=0.1262, simple_loss=0.2073, pruned_loss=0.0225, over 1248074.83 frames. ], batch size: 39, lr: 3.52e-03, grad_scale: 8.0 +2023-03-21 12:35:37,202 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=130316.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:35:56,784 INFO [train.py:901] (1/2) Epoch 47, batch 450, loss[loss=0.1236, simple_loss=0.2053, pruned_loss=0.02095, over 7261.00 frames. ], tot_loss[loss=0.1262, simple_loss=0.2078, pruned_loss=0.02233, over 1292964.95 frames. ], batch size: 52, lr: 3.52e-03, grad_scale: 8.0 +2023-03-21 12:35:57,374 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130356.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:35:57,452 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6036, 2.2785, 2.4403, 3.6447, 2.0126, 3.5012, 1.4593, 3.3817], + device='cuda:1'), covar=tensor([0.0238, 0.1663, 0.2083, 0.0336, 0.4047, 0.0344, 0.1464, 0.0485], + device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0242, 0.0255, 0.0213, 0.0250, 0.0219, 0.0220, 0.0229], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:35:58,364 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130358.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 12:36:01,349 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.257e+02 1.656e+02 1.997e+02 2.369e+02 3.103e+02, threshold=3.993e+02, percent-clipped=0.0 +2023-03-21 12:36:04,360 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 12:36:05,451 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 12:36:12,188 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130384.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:36:23,528 INFO [train.py:901] (1/2) Epoch 47, batch 500, loss[loss=0.1083, simple_loss=0.193, pruned_loss=0.01185, over 7344.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.208, pruned_loss=0.0227, over 1327657.67 frames. ], batch size: 44, lr: 3.52e-03, grad_scale: 8.0 +2023-03-21 12:36:29,507 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 +2023-03-21 12:36:37,420 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4130, 4.8866, 4.9481, 4.9031, 4.8207, 4.3996, 4.9625, 4.7909], + device='cuda:1'), covar=tensor([0.0453, 0.0405, 0.0393, 0.0436, 0.0330, 0.0433, 0.0350, 0.0431], + device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0272, 0.0212, 0.0208, 0.0164, 0.0239, 0.0218, 0.0153], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 12:36:37,482 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130432.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:36:37,845 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 12:36:38,901 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 12:36:39,876 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 12:36:40,491 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130438.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:36:41,875 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 12:36:44,082 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130445.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:36:45,061 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130447.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:36:46,484 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 12:36:48,949 INFO [train.py:901] (1/2) Epoch 47, batch 550, loss[loss=0.1494, simple_loss=0.2351, pruned_loss=0.03188, over 6794.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2081, pruned_loss=0.02266, over 1351534.79 frames. ], batch size: 106, lr: 3.52e-03, grad_scale: 8.0 +2023-03-21 12:36:54,108 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.358e+02 1.782e+02 2.105e+02 2.529e+02 4.728e+02, threshold=4.210e+02, percent-clipped=3.0 +2023-03-21 12:36:59,390 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 12:37:07,890 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 12:37:09,576 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130493.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:37:10,513 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=130495.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:37:10,992 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 12:37:12,655 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130499.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:37:15,524 INFO [train.py:901] (1/2) Epoch 47, batch 600, loss[loss=0.1266, simple_loss=0.2033, pruned_loss=0.02492, over 7254.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2079, pruned_loss=0.02243, over 1372134.13 frames. ], batch size: 55, lr: 3.52e-03, grad_scale: 8.0 +2023-03-21 12:37:18,044 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 12:37:34,331 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 12:37:35,447 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3757, 4.0070, 3.9786, 4.0083, 4.0399, 3.9348, 4.2158, 3.7401], + device='cuda:1'), covar=tensor([0.0179, 0.0165, 0.0132, 0.0167, 0.0393, 0.0114, 0.0138, 0.0192], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0105, 0.0107, 0.0093, 0.0182, 0.0113, 0.0109, 0.0118], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 12:37:38,137 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130548.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:37:42,157 INFO [train.py:901] (1/2) Epoch 47, batch 650, loss[loss=0.1272, simple_loss=0.2103, pruned_loss=0.02204, over 7106.00 frames. ], tot_loss[loss=0.1266, simple_loss=0.2081, pruned_loss=0.02256, over 1389291.58 frames. ], batch size: 98, lr: 3.52e-03, grad_scale: 8.0 +2023-03-21 12:37:43,702 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 12:37:46,756 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.631e+02 1.961e+02 2.425e+02 4.775e+02, threshold=3.922e+02, percent-clipped=1.0 +2023-03-21 12:37:47,342 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130565.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:38:00,818 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 12:38:07,800 INFO [train.py:901] (1/2) Epoch 47, batch 700, loss[loss=0.1382, simple_loss=0.2225, pruned_loss=0.02694, over 7256.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2078, pruned_loss=0.02246, over 1400366.58 frames. ], batch size: 64, lr: 3.52e-03, grad_scale: 8.0 +2023-03-21 12:38:09,939 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130609.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 12:38:10,325 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 12:38:18,408 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130626.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:38:34,370 INFO [train.py:901] (1/2) Epoch 47, batch 750, loss[loss=0.1378, simple_loss=0.2165, pruned_loss=0.02953, over 7212.00 frames. ], tot_loss[loss=0.1262, simple_loss=0.2076, pruned_loss=0.02243, over 1409744.35 frames. ], batch size: 50, lr: 3.52e-03, grad_scale: 8.0 +2023-03-21 12:38:34,978 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130656.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:38:35,357 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 12:38:35,883 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 12:38:35,970 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130658.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 12:38:38,843 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.113e+02 1.747e+02 2.015e+02 2.453e+02 3.447e+02, threshold=4.031e+02, percent-clipped=0.0 +2023-03-21 12:38:47,209 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.42 vs. limit=5.0 +2023-03-21 12:38:49,446 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 12:38:54,003 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 12:38:58,995 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=130704.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:38:59,433 INFO [train.py:901] (1/2) Epoch 47, batch 800, loss[loss=0.1179, simple_loss=0.2019, pruned_loss=0.01696, over 7329.00 frames. ], tot_loss[loss=0.1263, simple_loss=0.2076, pruned_loss=0.02246, over 1417369.72 frames. ], batch size: 75, lr: 3.52e-03, grad_scale: 8.0 +2023-03-21 12:38:59,987 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 12:39:00,025 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=130706.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:39:01,996 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 12:39:14,474 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 12:39:18,572 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130740.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:39:25,939 INFO [train.py:901] (1/2) Epoch 47, batch 850, loss[loss=0.1223, simple_loss=0.2089, pruned_loss=0.0178, over 7296.00 frames. ], tot_loss[loss=0.1262, simple_loss=0.2076, pruned_loss=0.02238, over 1422835.18 frames. ], batch size: 49, lr: 3.52e-03, grad_scale: 8.0 +2023-03-21 12:39:27,090 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9113, 4.0327, 3.8045, 3.9824, 3.6391, 3.8579, 4.1545, 4.1954], + device='cuda:1'), covar=tensor([0.0351, 0.0277, 0.0347, 0.0274, 0.0429, 0.0476, 0.0455, 0.0385], + device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0129, 0.0123, 0.0128, 0.0117, 0.0104, 0.0101, 0.0105], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 12:39:30,544 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.147e+02 1.714e+02 1.949e+02 2.244e+02 4.370e+02, threshold=3.898e+02, percent-clipped=1.0 +2023-03-21 12:39:32,147 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 12:39:32,156 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 12:39:37,576 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-03-21 12:39:37,714 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 12:39:38,842 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8588, 3.7038, 2.7433, 3.3977, 2.8241, 2.2615, 1.7474, 3.8130], + device='cuda:1'), covar=tensor([0.0058, 0.0085, 0.0235, 0.0101, 0.0226, 0.0666, 0.0765, 0.0061], + device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0096, 0.0117, 0.0101, 0.0136, 0.0136, 0.0130, 0.0108], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 12:39:40,740 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 12:39:42,776 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130788.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:39:45,816 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130794.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:39:52,265 INFO [train.py:901] (1/2) Epoch 47, batch 900, loss[loss=0.1124, simple_loss=0.1937, pruned_loss=0.01554, over 7154.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.208, pruned_loss=0.0224, over 1427609.18 frames. ], batch size: 41, lr: 3.51e-03, grad_scale: 8.0 +2023-03-21 12:40:02,354 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-21 12:40:18,166 INFO [train.py:901] (1/2) Epoch 47, batch 950, loss[loss=0.1322, simple_loss=0.2124, pruned_loss=0.02601, over 7320.00 frames. ], tot_loss[loss=0.1263, simple_loss=0.2078, pruned_loss=0.02239, over 1430268.65 frames. ], batch size: 49, lr: 3.51e-03, grad_scale: 8.0 +2023-03-21 12:40:19,152 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 12:40:22,691 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.245e+02 1.741e+02 1.966e+02 2.320e+02 4.605e+02, threshold=3.932e+02, percent-clipped=1.0 +2023-03-21 12:40:23,854 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1983, 2.5988, 2.0822, 2.8177, 2.9437, 2.8142, 2.3899, 2.5986], + device='cuda:1'), covar=tensor([0.2505, 0.1284, 0.3977, 0.0890, 0.0377, 0.0347, 0.0478, 0.0513], + device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0231, 0.0243, 0.0258, 0.0202, 0.0205, 0.0221, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:40:29,844 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4096, 4.8542, 4.9045, 4.8915, 4.7950, 4.4029, 4.9621, 4.7923], + device='cuda:1'), covar=tensor([0.0408, 0.0376, 0.0373, 0.0399, 0.0330, 0.0401, 0.0333, 0.0435], + device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0272, 0.0210, 0.0207, 0.0163, 0.0237, 0.0218, 0.0152], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 12:40:39,774 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1173, 3.6065, 2.5481, 3.7188, 2.8023, 3.5531, 1.7019, 2.4448], + device='cuda:1'), covar=tensor([0.0414, 0.0842, 0.3120, 0.0687, 0.0457, 0.0554, 0.4293, 0.2172], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0255, 0.0277, 0.0267, 0.0263, 0.0259, 0.0227, 0.0253], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:40:42,275 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 12:40:44,418 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130904.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 12:40:44,847 INFO [train.py:901] (1/2) Epoch 47, batch 1000, loss[loss=0.1204, simple_loss=0.2055, pruned_loss=0.01769, over 7362.00 frames. ], tot_loss[loss=0.1265, simple_loss=0.2082, pruned_loss=0.02239, over 1433823.38 frames. ], batch size: 73, lr: 3.51e-03, grad_scale: 8.0 +2023-03-21 12:40:53,126 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130921.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:41:02,626 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 12:41:10,218 INFO [train.py:901] (1/2) Epoch 47, batch 1050, loss[loss=0.1446, simple_loss=0.2239, pruned_loss=0.03264, over 7371.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.208, pruned_loss=0.02239, over 1435614.49 frames. ], batch size: 65, lr: 3.51e-03, grad_scale: 8.0 +2023-03-21 12:41:14,746 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.118e+02 1.723e+02 1.973e+02 2.315e+02 3.318e+02, threshold=3.946e+02, percent-clipped=0.0 +2023-03-21 12:41:25,005 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 12:41:29,065 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 12:41:37,121 INFO [train.py:901] (1/2) Epoch 47, batch 1100, loss[loss=0.1307, simple_loss=0.2093, pruned_loss=0.02609, over 7246.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.208, pruned_loss=0.02241, over 1436714.35 frames. ], batch size: 55, lr: 3.51e-03, grad_scale: 8.0 +2023-03-21 12:41:54,980 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131040.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:41:57,958 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 12:41:58,483 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 12:42:02,449 INFO [train.py:901] (1/2) Epoch 47, batch 1150, loss[loss=0.1453, simple_loss=0.2265, pruned_loss=0.03206, over 7253.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.2074, pruned_loss=0.02223, over 1437451.63 frames. ], batch size: 64, lr: 3.51e-03, grad_scale: 8.0 +2023-03-21 12:42:07,684 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.315e+02 1.713e+02 2.076e+02 2.452e+02 6.009e+02, threshold=4.153e+02, percent-clipped=1.0 +2023-03-21 12:42:11,903 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 12:42:12,421 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 12:42:13,540 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6039, 3.5942, 2.4694, 3.2654, 2.5907, 2.3650, 1.9748, 3.6444], + device='cuda:1'), covar=tensor([0.0065, 0.0059, 0.0264, 0.0122, 0.0263, 0.0555, 0.0709, 0.0066], + device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0096, 0.0118, 0.0101, 0.0136, 0.0137, 0.0131, 0.0109], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 12:42:14,350 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 +2023-03-21 12:42:20,535 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:42:20,586 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:42:22,170 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6892, 1.7330, 2.2072, 2.3909, 2.1680, 2.3951, 2.0696, 2.3296], + device='cuda:1'), covar=tensor([0.3754, 0.4030, 0.1162, 0.0736, 0.2157, 0.1413, 0.1682, 0.3393], + device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0086, 0.0079, 0.0070, 0.0071, 0.0069, 0.0110, 0.0072], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 12:42:23,674 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131094.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:42:29,220 INFO [train.py:901] (1/2) Epoch 47, batch 1200, loss[loss=0.1305, simple_loss=0.2166, pruned_loss=0.02224, over 7323.00 frames. ], tot_loss[loss=0.1258, simple_loss=0.207, pruned_loss=0.02236, over 1435937.28 frames. ], batch size: 61, lr: 3.51e-03, grad_scale: 8.0 +2023-03-21 12:42:43,925 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 12:42:44,997 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=131136.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:42:48,007 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=131142.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:42:50,330 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 +2023-03-21 12:42:55,881 INFO [train.py:901] (1/2) Epoch 47, batch 1250, loss[loss=0.1361, simple_loss=0.2229, pruned_loss=0.02469, over 7294.00 frames. ], tot_loss[loss=0.1263, simple_loss=0.2073, pruned_loss=0.02266, over 1436446.37 frames. ], batch size: 66, lr: 3.51e-03, grad_scale: 8.0 +2023-03-21 12:43:00,376 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.326e+02 1.727e+02 1.988e+02 2.304e+02 5.808e+02, threshold=3.975e+02, percent-clipped=1.0 +2023-03-21 12:43:05,755 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131174.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:43:09,175 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 12:43:13,294 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 12:43:14,807 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 12:43:15,981 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8483, 2.1579, 2.2224, 2.7007, 2.5156, 2.6083, 2.5858, 2.5742], + device='cuda:1'), covar=tensor([0.2500, 0.3721, 0.1467, 0.1879, 0.1860, 0.2418, 0.1938, 0.2243], + device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0087, 0.0080, 0.0070, 0.0072, 0.0070, 0.0112, 0.0073], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 12:43:21,456 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131204.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 12:43:21,863 INFO [train.py:901] (1/2) Epoch 47, batch 1300, loss[loss=0.1297, simple_loss=0.2061, pruned_loss=0.02659, over 7322.00 frames. ], tot_loss[loss=0.1261, simple_loss=0.2071, pruned_loss=0.02252, over 1437481.41 frames. ], batch size: 61, lr: 3.51e-03, grad_scale: 8.0 +2023-03-21 12:43:30,152 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131221.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:43:32,171 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7447, 3.1055, 2.6248, 2.9114, 2.9566, 2.7285, 3.0158, 2.7078], + device='cuda:1'), covar=tensor([0.1116, 0.0493, 0.1103, 0.0959, 0.0677, 0.0693, 0.0985, 0.1138], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0062, 0.0071, 0.0061, 0.0059, 0.0066, 0.0059, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:43:37,774 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131235.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 12:43:38,151 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 12:43:41,316 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 12:43:44,453 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 12:43:47,082 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=131252.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:43:48,580 INFO [train.py:901] (1/2) Epoch 47, batch 1350, loss[loss=0.113, simple_loss=0.1943, pruned_loss=0.01583, over 7332.00 frames. ], tot_loss[loss=0.1254, simple_loss=0.2068, pruned_loss=0.02203, over 1436756.65 frames. ], batch size: 44, lr: 3.51e-03, grad_scale: 8.0 +2023-03-21 12:43:53,125 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.291e+02 1.774e+02 1.993e+02 2.320e+02 3.599e+02, threshold=3.986e+02, percent-clipped=0.0 +2023-03-21 12:43:53,653 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 12:43:55,644 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=131269.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:43:59,203 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1933, 4.6955, 4.7709, 4.7434, 4.6776, 4.2494, 4.7850, 4.6118], + device='cuda:1'), covar=tensor([0.0492, 0.0437, 0.0402, 0.0466, 0.0334, 0.0429, 0.0367, 0.0416], + device='cuda:1'), in_proj_covar=tensor([0.0220, 0.0270, 0.0209, 0.0206, 0.0162, 0.0237, 0.0218, 0.0151], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 12:44:13,897 INFO [train.py:901] (1/2) Epoch 47, batch 1400, loss[loss=0.1286, simple_loss=0.2152, pruned_loss=0.02105, over 7274.00 frames. ], tot_loss[loss=0.1258, simple_loss=0.2074, pruned_loss=0.02209, over 1439779.66 frames. ], batch size: 55, lr: 3.51e-03, grad_scale: 8.0 +2023-03-21 12:44:27,251 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 12:44:40,388 INFO [train.py:901] (1/2) Epoch 47, batch 1450, loss[loss=0.1239, simple_loss=0.207, pruned_loss=0.02045, over 7246.00 frames. ], tot_loss[loss=0.1255, simple_loss=0.2071, pruned_loss=0.02199, over 1438432.99 frames. ], batch size: 55, lr: 3.51e-03, grad_scale: 8.0 +2023-03-21 12:44:45,044 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.405e+02 1.823e+02 2.131e+02 2.524e+02 5.111e+02, threshold=4.261e+02, percent-clipped=2.0 +2023-03-21 12:44:51,653 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 12:45:06,093 INFO [train.py:901] (1/2) Epoch 47, batch 1500, loss[loss=0.1067, simple_loss=0.1851, pruned_loss=0.0142, over 7328.00 frames. ], tot_loss[loss=0.1262, simple_loss=0.2074, pruned_loss=0.02254, over 1437528.26 frames. ], batch size: 44, lr: 3.51e-03, grad_scale: 8.0 +2023-03-21 12:45:08,834 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 12:45:08,966 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131409.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:45:12,870 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-03-21 12:45:13,179 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131416.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:45:32,712 INFO [train.py:901] (1/2) Epoch 47, batch 1550, loss[loss=0.1305, simple_loss=0.2126, pruned_loss=0.02422, over 7211.00 frames. ], tot_loss[loss=0.1266, simple_loss=0.2078, pruned_loss=0.02266, over 1438229.44 frames. ], batch size: 93, lr: 3.51e-03, grad_scale: 8.0 +2023-03-21 12:45:33,714 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 12:45:37,221 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.357e+02 1.828e+02 2.189e+02 2.514e+02 5.463e+02, threshold=4.378e+02, percent-clipped=3.0 +2023-03-21 12:45:40,399 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131470.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:45:43,948 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131477.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:45:51,493 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131492.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:45:59,268 INFO [train.py:901] (1/2) Epoch 47, batch 1600, loss[loss=0.1314, simple_loss=0.2231, pruned_loss=0.01979, over 6685.00 frames. ], tot_loss[loss=0.127, simple_loss=0.2081, pruned_loss=0.02297, over 1437147.75 frames. ], batch size: 106, lr: 3.51e-03, grad_scale: 8.0 +2023-03-21 12:46:04,806 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 12:46:05,324 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 12:46:08,411 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 12:46:11,999 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131530.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 12:46:18,027 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 12:46:22,184 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4666, 3.7299, 2.6814, 4.0492, 3.1987, 3.4296, 1.8651, 2.8761], + device='cuda:1'), covar=tensor([0.0545, 0.0873, 0.2773, 0.0536, 0.0548, 0.0970, 0.4426, 0.1829], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0253, 0.0273, 0.0264, 0.0261, 0.0257, 0.0226, 0.0251], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:46:22,535 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 12:46:23,706 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131553.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:46:24,577 INFO [train.py:901] (1/2) Epoch 47, batch 1650, loss[loss=0.1237, simple_loss=0.2097, pruned_loss=0.01881, over 7325.00 frames. ], tot_loss[loss=0.1273, simple_loss=0.2088, pruned_loss=0.02296, over 1441068.53 frames. ], batch size: 59, lr: 3.50e-03, grad_scale: 8.0 +2023-03-21 12:46:28,967 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.299e+02 1.715e+02 1.954e+02 2.297e+02 4.691e+02, threshold=3.908e+02, percent-clipped=1.0 +2023-03-21 12:46:29,969 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 12:46:47,884 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 12:46:51,289 INFO [train.py:901] (1/2) Epoch 47, batch 1700, loss[loss=0.1153, simple_loss=0.2003, pruned_loss=0.01508, over 7253.00 frames. ], tot_loss[loss=0.1276, simple_loss=0.2092, pruned_loss=0.023, over 1442105.15 frames. ], batch size: 89, lr: 3.50e-03, grad_scale: 8.0 +2023-03-21 12:46:52,349 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 12:47:02,891 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 12:47:05,736 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-03-21 12:47:16,418 INFO [train.py:901] (1/2) Epoch 47, batch 1750, loss[loss=0.1211, simple_loss=0.206, pruned_loss=0.01815, over 7268.00 frames. ], tot_loss[loss=0.1272, simple_loss=0.2086, pruned_loss=0.02287, over 1441945.01 frames. ], batch size: 66, lr: 3.50e-03, grad_scale: 8.0 +2023-03-21 12:47:20,891 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.778e+02 2.043e+02 2.310e+02 3.986e+02, threshold=4.086e+02, percent-clipped=1.0 +2023-03-21 12:47:28,167 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 12:47:29,154 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 12:47:39,430 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6767, 1.7221, 1.7815, 2.0653, 2.0136, 2.0821, 1.6135, 2.1349], + device='cuda:1'), covar=tensor([0.3304, 0.3877, 0.2085, 0.1419, 0.2747, 0.2077, 0.1726, 0.1541], + device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0087, 0.0079, 0.0070, 0.0071, 0.0071, 0.0112, 0.0073], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 12:47:42,810 INFO [train.py:901] (1/2) Epoch 47, batch 1800, loss[loss=0.1383, simple_loss=0.2219, pruned_loss=0.02737, over 6717.00 frames. ], tot_loss[loss=0.1269, simple_loss=0.2085, pruned_loss=0.02271, over 1441600.71 frames. ], batch size: 106, lr: 3.50e-03, grad_scale: 8.0 +2023-03-21 12:47:49,952 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 12:48:03,404 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 12:48:07,910 INFO [train.py:901] (1/2) Epoch 47, batch 1850, loss[loss=0.1287, simple_loss=0.2106, pruned_loss=0.02343, over 7355.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2082, pruned_loss=0.02261, over 1441884.83 frames. ], batch size: 63, lr: 3.50e-03, grad_scale: 16.0 +2023-03-21 12:48:12,871 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 +2023-03-21 12:48:13,549 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.241e+02 1.650e+02 1.902e+02 2.256e+02 3.427e+02, threshold=3.803e+02, percent-clipped=0.0 +2023-03-21 12:48:14,163 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131765.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:48:14,637 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 12:48:17,768 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131772.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:48:30,073 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1132, 2.4562, 1.8555, 3.0929, 3.0016, 2.7100, 2.5993, 2.5695], + device='cuda:1'), covar=tensor([0.2443, 0.1246, 0.4268, 0.0751, 0.0400, 0.0373, 0.0484, 0.0457], + device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0231, 0.0244, 0.0257, 0.0202, 0.0205, 0.0221, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:48:30,420 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 12:48:34,813 INFO [train.py:901] (1/2) Epoch 47, batch 1900, loss[loss=0.1371, simple_loss=0.2134, pruned_loss=0.03045, over 7312.00 frames. ], tot_loss[loss=0.127, simple_loss=0.2088, pruned_loss=0.02262, over 1444266.28 frames. ], batch size: 59, lr: 3.50e-03, grad_scale: 16.0 +2023-03-21 12:48:38,130 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-03-21 12:48:47,659 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131830.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:48:48,172 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.4329, 1.6525, 1.4566, 1.6703, 1.6829, 1.7227, 1.5194, 1.3662], + device='cuda:1'), covar=tensor([0.0199, 0.0211, 0.0334, 0.0212, 0.0159, 0.0147, 0.0217, 0.0266], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0042, 0.0039, 0.0042, 0.0039, 0.0038, 0.0040, 0.0050], + device='cuda:1'), out_proj_covar=tensor([4.7670e-05, 4.6724e-05, 4.4370e-05, 4.5947e-05, 4.3401e-05, 4.2250e-05, + 4.5409e-05, 5.4795e-05], device='cuda:1') +2023-03-21 12:48:56,795 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 12:48:57,361 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131848.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:49:01,452 INFO [train.py:901] (1/2) Epoch 47, batch 1950, loss[loss=0.1399, simple_loss=0.221, pruned_loss=0.02934, over 7295.00 frames. ], tot_loss[loss=0.1273, simple_loss=0.2092, pruned_loss=0.02268, over 1446760.70 frames. ], batch size: 86, lr: 3.50e-03, grad_scale: 8.0 +2023-03-21 12:49:02,646 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131857.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:49:06,462 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.255e+02 1.657e+02 1.977e+02 2.340e+02 3.243e+02, threshold=3.954e+02, percent-clipped=0.0 +2023-03-21 12:49:08,001 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 12:49:13,140 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 12:49:13,191 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=131878.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:49:13,648 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 12:49:26,696 INFO [train.py:901] (1/2) Epoch 47, batch 2000, loss[loss=0.1318, simple_loss=0.2228, pruned_loss=0.02043, over 6796.00 frames. ], tot_loss[loss=0.1271, simple_loss=0.2088, pruned_loss=0.02266, over 1446531.29 frames. ], batch size: 107, lr: 3.50e-03, grad_scale: 8.0 +2023-03-21 12:49:29,253 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 12:49:33,345 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131918.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:49:41,376 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 12:49:50,828 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 12:49:53,417 INFO [train.py:901] (1/2) Epoch 47, batch 2050, loss[loss=0.1273, simple_loss=0.2054, pruned_loss=0.02462, over 7272.00 frames. ], tot_loss[loss=0.127, simple_loss=0.2085, pruned_loss=0.02271, over 1445131.49 frames. ], batch size: 47, lr: 3.50e-03, grad_scale: 8.0 +2023-03-21 12:49:55,235 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 +2023-03-21 12:49:58,473 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 1.826e+02 2.096e+02 2.396e+02 3.231e+02, threshold=4.192e+02, percent-clipped=0.0 +2023-03-21 12:50:04,719 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0556, 2.1141, 2.2839, 3.3610, 1.8482, 3.1192, 1.3579, 3.0298], + device='cuda:1'), covar=tensor([0.0260, 0.1821, 0.2165, 0.0290, 0.4467, 0.0343, 0.1467, 0.0643], + device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0243, 0.0254, 0.0211, 0.0247, 0.0219, 0.0220, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:50:05,203 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4871, 2.4608, 2.6141, 3.7494, 2.0492, 3.5660, 1.6137, 3.2500], + device='cuda:1'), covar=tensor([0.0201, 0.1712, 0.2046, 0.0225, 0.4304, 0.0317, 0.1468, 0.0537], + device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0243, 0.0254, 0.0211, 0.0247, 0.0219, 0.0220, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:50:22,454 INFO [train.py:901] (1/2) Epoch 47, batch 2100, loss[loss=0.1344, simple_loss=0.2102, pruned_loss=0.02934, over 7236.00 frames. ], tot_loss[loss=0.1273, simple_loss=0.209, pruned_loss=0.02277, over 1445459.89 frames. ], batch size: 50, lr: 3.50e-03, grad_scale: 8.0 +2023-03-21 12:50:27,467 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 12:50:28,127 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4165, 2.2550, 2.6110, 2.2921, 2.5686, 2.5018, 2.2571, 1.9164], + device='cuda:1'), covar=tensor([0.0390, 0.0547, 0.0279, 0.0292, 0.0538, 0.0335, 0.0400, 0.0270], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0042, 0.0043, 0.0042, 0.0040, 0.0041, 0.0046, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 12:50:31,075 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 12:50:36,838 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132031.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:50:48,447 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132054.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:50:48,798 INFO [train.py:901] (1/2) Epoch 47, batch 2150, loss[loss=0.1197, simple_loss=0.2018, pruned_loss=0.01881, over 7262.00 frames. ], tot_loss[loss=0.1269, simple_loss=0.2087, pruned_loss=0.0225, over 1446986.22 frames. ], batch size: 70, lr: 3.50e-03, grad_scale: 8.0 +2023-03-21 12:50:53,614 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 +2023-03-21 12:50:53,763 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.210e+02 1.654e+02 1.954e+02 2.342e+02 4.874e+02, threshold=3.908e+02, percent-clipped=1.0 +2023-03-21 12:50:53,903 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132065.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:50:57,476 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132072.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:51:07,553 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132092.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:51:09,704 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 +2023-03-21 12:51:13,869 INFO [train.py:901] (1/2) Epoch 47, batch 2200, loss[loss=0.1369, simple_loss=0.215, pruned_loss=0.02942, over 7263.00 frames. ], tot_loss[loss=0.1271, simple_loss=0.209, pruned_loss=0.02267, over 1446957.55 frames. ], batch size: 64, lr: 3.50e-03, grad_scale: 8.0 +2023-03-21 12:51:18,681 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 12:51:19,219 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=132113.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:51:20,285 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132115.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 12:51:22,723 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=132120.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:51:23,325 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8559, 4.0200, 3.7468, 3.9660, 3.6350, 3.9474, 4.2934, 4.2730], + device='cuda:1'), covar=tensor([0.0221, 0.0158, 0.0270, 0.0181, 0.0348, 0.0270, 0.0205, 0.0174], + device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0130, 0.0124, 0.0129, 0.0117, 0.0105, 0.0101, 0.0105], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 12:51:23,834 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5380, 3.6742, 3.4258, 3.6179, 3.3694, 3.5541, 3.9144, 3.9173], + device='cuda:1'), covar=tensor([0.0252, 0.0189, 0.0276, 0.0218, 0.0363, 0.0448, 0.0220, 0.0193], + device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0130, 0.0124, 0.0129, 0.0117, 0.0105, 0.0101, 0.0105], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 12:51:36,948 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132148.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:51:40,433 INFO [train.py:901] (1/2) Epoch 47, batch 2250, loss[loss=0.1205, simple_loss=0.1958, pruned_loss=0.0226, over 7204.00 frames. ], tot_loss[loss=0.1271, simple_loss=0.2088, pruned_loss=0.02276, over 1446754.18 frames. ], batch size: 45, lr: 3.50e-03, grad_scale: 8.0 +2023-03-21 12:51:43,351 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 +2023-03-21 12:51:45,421 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.229e+02 1.779e+02 2.110e+02 2.462e+02 5.777e+02, threshold=4.221e+02, percent-clipped=3.0 +2023-03-21 12:51:51,593 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 12:51:52,082 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 12:52:02,422 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=132196.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:52:02,501 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3603, 2.3975, 2.6513, 2.3795, 2.5872, 2.6915, 2.3626, 2.1372], + device='cuda:1'), covar=tensor([0.0728, 0.0478, 0.0319, 0.0333, 0.0558, 0.0318, 0.0349, 0.0340], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0042, 0.0043, 0.0043, 0.0040, 0.0041, 0.0046, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 12:52:05,620 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 12:52:07,163 INFO [train.py:901] (1/2) Epoch 47, batch 2300, loss[loss=0.1077, simple_loss=0.186, pruned_loss=0.0147, over 7183.00 frames. ], tot_loss[loss=0.127, simple_loss=0.2086, pruned_loss=0.02268, over 1447065.24 frames. ], batch size: 39, lr: 3.50e-03, grad_scale: 8.0 +2023-03-21 12:52:11,258 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132213.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:52:18,366 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4784, 2.4007, 2.6027, 2.3816, 2.5397, 2.5348, 2.3324, 1.9958], + device='cuda:1'), covar=tensor([0.0416, 0.0395, 0.0344, 0.0300, 0.0498, 0.0419, 0.0291, 0.0273], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0042, 0.0043, 0.0043, 0.0040, 0.0041, 0.0046, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 12:52:19,428 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3203, 3.2195, 3.1904, 3.3904, 3.0193, 2.8607, 3.5063, 2.3487], + device='cuda:1'), covar=tensor([0.0651, 0.0791, 0.0945, 0.0939, 0.0857, 0.1212, 0.0997, 0.3102], + device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0337, 0.0268, 0.0348, 0.0277, 0.0285, 0.0342, 0.0236], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 12:52:23,519 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-03-21 12:52:32,353 INFO [train.py:901] (1/2) Epoch 47, batch 2350, loss[loss=0.1334, simple_loss=0.2066, pruned_loss=0.03007, over 7272.00 frames. ], tot_loss[loss=0.1265, simple_loss=0.2081, pruned_loss=0.02245, over 1445938.43 frames. ], batch size: 52, lr: 3.50e-03, grad_scale: 8.0 +2023-03-21 12:52:37,445 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.200e+02 1.770e+02 2.102e+02 2.451e+02 5.297e+02, threshold=4.204e+02, percent-clipped=1.0 +2023-03-21 12:52:38,599 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6876, 1.5661, 1.8461, 2.0785, 1.9543, 2.1033, 1.5180, 2.1714], + device='cuda:1'), covar=tensor([0.2616, 0.4106, 0.1610, 0.1485, 0.1803, 0.2144, 0.1925, 0.1613], + device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0085, 0.0078, 0.0069, 0.0070, 0.0070, 0.0110, 0.0071], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 12:52:52,486 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 12:52:58,524 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 12:52:59,021 INFO [train.py:901] (1/2) Epoch 47, batch 2400, loss[loss=0.1376, simple_loss=0.2211, pruned_loss=0.02709, over 7342.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2079, pruned_loss=0.02241, over 1444714.46 frames. ], batch size: 54, lr: 3.49e-03, grad_scale: 8.0 +2023-03-21 12:53:09,195 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 12:53:12,251 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 12:53:24,325 INFO [train.py:901] (1/2) Epoch 47, batch 2450, loss[loss=0.1378, simple_loss=0.2205, pruned_loss=0.02754, over 7277.00 frames. ], tot_loss[loss=0.1268, simple_loss=0.208, pruned_loss=0.02283, over 1445541.01 frames. ], batch size: 77, lr: 3.49e-03, grad_scale: 8.0 +2023-03-21 12:53:29,373 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.194e+02 1.852e+02 2.168e+02 2.561e+02 3.828e+02, threshold=4.336e+02, percent-clipped=0.0 +2023-03-21 12:53:39,263 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 12:53:41,732 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132387.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:53:46,460 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6449, 3.4402, 3.5227, 3.7297, 3.4482, 3.1808, 3.9022, 2.5965], + device='cuda:1'), covar=tensor([0.0531, 0.0678, 0.0892, 0.0804, 0.0789, 0.1112, 0.0943, 0.3070], + device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0338, 0.0269, 0.0349, 0.0278, 0.0286, 0.0343, 0.0237], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 12:53:51,201 INFO [train.py:901] (1/2) Epoch 47, batch 2500, loss[loss=0.1188, simple_loss=0.2019, pruned_loss=0.01785, over 7363.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2079, pruned_loss=0.02273, over 1444657.83 frames. ], batch size: 73, lr: 3.49e-03, grad_scale: 8.0 +2023-03-21 12:53:53,840 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132410.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 12:54:04,884 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 12:54:06,152 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-03-21 12:54:17,076 INFO [train.py:901] (1/2) Epoch 47, batch 2550, loss[loss=0.1292, simple_loss=0.2234, pruned_loss=0.01748, over 7219.00 frames. ], tot_loss[loss=0.1266, simple_loss=0.2079, pruned_loss=0.02263, over 1445005.64 frames. ], batch size: 93, lr: 3.49e-03, grad_scale: 8.0 +2023-03-21 12:54:22,811 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.239e+02 1.683e+02 1.964e+02 2.400e+02 4.092e+02, threshold=3.929e+02, percent-clipped=0.0 +2023-03-21 12:54:36,625 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3807, 2.3175, 2.5082, 3.6016, 2.0757, 3.3838, 1.5946, 3.2022], + device='cuda:1'), covar=tensor([0.0196, 0.1586, 0.1898, 0.0236, 0.4190, 0.0287, 0.1492, 0.0407], + device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0242, 0.0254, 0.0210, 0.0246, 0.0218, 0.0219, 0.0229], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:54:43,035 INFO [train.py:901] (1/2) Epoch 47, batch 2600, loss[loss=0.11, simple_loss=0.1971, pruned_loss=0.01145, over 7343.00 frames. ], tot_loss[loss=0.1268, simple_loss=0.2082, pruned_loss=0.0227, over 1443039.60 frames. ], batch size: 44, lr: 3.49e-03, grad_scale: 8.0 +2023-03-21 12:54:47,120 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132513.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:55:08,062 INFO [train.py:901] (1/2) Epoch 47, batch 2650, loss[loss=0.1266, simple_loss=0.2103, pruned_loss=0.02144, over 7308.00 frames. ], tot_loss[loss=0.1263, simple_loss=0.208, pruned_loss=0.02236, over 1440728.20 frames. ], batch size: 59, lr: 3.49e-03, grad_scale: 8.0 +2023-03-21 12:55:11,140 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=132561.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:55:11,406 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 +2023-03-21 12:55:13,051 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.743e+02 2.078e+02 2.512e+02 4.321e+02, threshold=4.155e+02, percent-clipped=2.0 +2023-03-21 12:55:17,603 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.7983, 4.2160, 4.4381, 4.3979, 4.4806, 4.3555, 4.6842, 4.2408], + device='cuda:1'), covar=tensor([0.0167, 0.0177, 0.0121, 0.0169, 0.0345, 0.0107, 0.0116, 0.0131], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0106, 0.0106, 0.0094, 0.0183, 0.0113, 0.0109, 0.0117], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 12:55:33,181 INFO [train.py:901] (1/2) Epoch 47, batch 2700, loss[loss=0.1188, simple_loss=0.2023, pruned_loss=0.01763, over 7314.00 frames. ], tot_loss[loss=0.126, simple_loss=0.2075, pruned_loss=0.02227, over 1441849.58 frames. ], batch size: 59, lr: 3.49e-03, grad_scale: 8.0 +2023-03-21 12:55:35,590 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-03-21 12:55:42,410 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 +2023-03-21 12:55:59,140 INFO [train.py:901] (1/2) Epoch 47, batch 2750, loss[loss=0.1405, simple_loss=0.221, pruned_loss=0.03004, over 7330.00 frames. ], tot_loss[loss=0.1263, simple_loss=0.208, pruned_loss=0.02232, over 1444131.33 frames. ], batch size: 59, lr: 3.49e-03, grad_scale: 8.0 +2023-03-21 12:56:01,801 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132660.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:56:04,149 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.232e+02 1.754e+02 2.060e+02 2.369e+02 4.641e+02, threshold=4.120e+02, percent-clipped=1.0 +2023-03-21 12:56:15,152 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132687.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:56:18,570 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5088, 2.6582, 3.4781, 3.4568, 3.5202, 3.5784, 3.3076, 3.4289], + device='cuda:1'), covar=tensor([0.0030, 0.0158, 0.0037, 0.0031, 0.0031, 0.0029, 0.0080, 0.0051], + device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0073, 0.0061, 0.0059, 0.0056, 0.0061, 0.0049, 0.0081], + device='cuda:1'), out_proj_covar=tensor([8.3816e-05, 1.4341e-04, 1.0704e-04, 9.8299e-05, 9.0915e-05, 1.0395e-04, + 8.9941e-05, 1.4514e-04], device='cuda:1') +2023-03-21 12:56:23,843 INFO [train.py:901] (1/2) Epoch 47, batch 2800, loss[loss=0.1376, simple_loss=0.2196, pruned_loss=0.02781, over 7331.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2089, pruned_loss=0.02225, over 1443554.63 frames. ], batch size: 61, lr: 3.49e-03, grad_scale: 8.0 +2023-03-21 12:56:26,411 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132710.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 12:56:31,916 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132721.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:56:46,873 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 12:56:46,933 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 12:56:46,958 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 12:56:46,977 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 12:56:47,158 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 12:56:47,174 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 12:56:47,187 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 12:56:47,292 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 12:56:47,370 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 12:56:47,385 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 12:56:47,403 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 12:56:47,573 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 12:56:47,578 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 12:56:47,587 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 12:56:48,075 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 12:56:48,078 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 12:56:48,141 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 12:56:53,002 INFO [train.py:901] (1/2) Epoch 48, batch 0, loss[loss=0.1327, simple_loss=0.2176, pruned_loss=0.0239, over 7217.00 frames. ], tot_loss[loss=0.1327, simple_loss=0.2176, pruned_loss=0.0239, over 7217.00 frames. ], batch size: 93, lr: 3.45e-03, grad_scale: 8.0 +2023-03-21 12:56:53,002 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 12:57:19,428 INFO [train.py:935] (1/2) Epoch 48, validation: loss=0.1658, simple_loss=0.2588, pruned_loss=0.03638, over 1622729.00 frames. +2023-03-21 12:57:19,429 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 12:57:22,411 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=132735.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:57:25,872 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 12:57:28,050 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132746.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:57:34,020 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=132758.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:57:36,958 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 12:57:37,422 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 1.668e+02 1.991e+02 2.294e+02 3.667e+02, threshold=3.982e+02, percent-clipped=0.0 +2023-03-21 12:57:40,150 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5883, 3.0360, 2.2973, 3.4477, 3.4403, 3.4854, 2.8926, 2.9442], + device='cuda:1'), covar=tensor([0.2297, 0.1125, 0.4194, 0.0839, 0.0350, 0.0448, 0.0574, 0.0775], + device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0234, 0.0249, 0.0261, 0.0205, 0.0208, 0.0224, 0.0231], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:57:44,538 INFO [train.py:901] (1/2) Epoch 48, batch 50, loss[loss=0.1264, simple_loss=0.2098, pruned_loss=0.02153, over 7248.00 frames. ], tot_loss[loss=0.1252, simple_loss=0.2062, pruned_loss=0.02209, over 324526.44 frames. ], batch size: 55, lr: 3.45e-03, grad_scale: 8.0 +2023-03-21 12:57:44,576 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 12:57:47,103 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 12:57:49,697 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 12:57:57,198 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5987, 2.5761, 2.6643, 2.5368, 2.7601, 2.7500, 2.6001, 2.1845], + device='cuda:1'), covar=tensor([0.0691, 0.0558, 0.0394, 0.0502, 0.0740, 0.0392, 0.0391, 0.0512], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0042, 0.0044, 0.0042, 0.0040, 0.0041, 0.0046, 0.0046], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 12:57:59,238 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132807.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:58:06,771 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 12:58:07,745 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 12:58:10,791 INFO [train.py:901] (1/2) Epoch 48, batch 100, loss[loss=0.1173, simple_loss=0.1969, pruned_loss=0.01882, over 7346.00 frames. ], tot_loss[loss=0.1252, simple_loss=0.2069, pruned_loss=0.02171, over 573413.44 frames. ], batch size: 54, lr: 3.45e-03, grad_scale: 8.0 +2023-03-21 12:58:28,997 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.188e+02 1.757e+02 1.926e+02 2.244e+02 3.383e+02, threshold=3.853e+02, percent-clipped=0.0 +2023-03-21 12:58:36,314 INFO [train.py:901] (1/2) Epoch 48, batch 150, loss[loss=0.1384, simple_loss=0.2194, pruned_loss=0.02868, over 7286.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2077, pruned_loss=0.02257, over 766737.33 frames. ], batch size: 57, lr: 3.45e-03, grad_scale: 8.0 +2023-03-21 12:58:43,459 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.3716, 4.8488, 4.6838, 5.3959, 5.1565, 5.2648, 4.7219, 4.9234], + device='cuda:1'), covar=tensor([0.0881, 0.2620, 0.2277, 0.0938, 0.0912, 0.1139, 0.0675, 0.1108], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0404, 0.0305, 0.0317, 0.0235, 0.0379, 0.0240, 0.0288], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 12:58:43,795 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 +2023-03-21 12:59:02,769 INFO [train.py:901] (1/2) Epoch 48, batch 200, loss[loss=0.1381, simple_loss=0.2191, pruned_loss=0.0286, over 7268.00 frames. ], tot_loss[loss=0.1256, simple_loss=0.2066, pruned_loss=0.02226, over 915318.16 frames. ], batch size: 55, lr: 3.45e-03, grad_scale: 8.0 +2023-03-21 12:59:09,903 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 12:59:13,927 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 12:59:20,387 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 12:59:20,849 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.260e+02 1.731e+02 2.035e+02 2.345e+02 3.653e+02, threshold=4.069e+02, percent-clipped=0.0 +2023-03-21 12:59:27,888 INFO [train.py:901] (1/2) Epoch 48, batch 250, loss[loss=0.151, simple_loss=0.2305, pruned_loss=0.03571, over 6694.00 frames. ], tot_loss[loss=0.1261, simple_loss=0.2071, pruned_loss=0.02258, over 1031154.88 frames. ], batch size: 106, lr: 3.45e-03, grad_scale: 8.0 +2023-03-21 12:59:30,233 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-03-21 12:59:32,917 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 12:59:38,168 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5336, 2.8213, 2.5111, 2.6784, 2.7393, 2.5415, 2.6542, 2.5372], + device='cuda:1'), covar=tensor([0.0715, 0.0672, 0.1060, 0.0787, 0.0797, 0.0742, 0.0946, 0.0820], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0062, 0.0071, 0.0063, 0.0059, 0.0066, 0.0060, 0.0057], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 12:59:48,092 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133016.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 12:59:50,607 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9180, 4.0532, 3.8136, 4.0068, 3.6622, 4.0073, 4.2849, 4.2770], + device='cuda:1'), covar=tensor([0.0202, 0.0153, 0.0227, 0.0187, 0.0309, 0.0271, 0.0261, 0.0235], + device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0129, 0.0123, 0.0128, 0.0116, 0.0103, 0.0101, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 12:59:54,512 INFO [train.py:901] (1/2) Epoch 48, batch 300, loss[loss=0.1245, simple_loss=0.2072, pruned_loss=0.02088, over 7263.00 frames. ], tot_loss[loss=0.1261, simple_loss=0.2073, pruned_loss=0.02244, over 1123055.01 frames. ], batch size: 64, lr: 3.45e-03, grad_scale: 8.0 +2023-03-21 12:59:54,524 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 13:00:03,625 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 13:00:12,665 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.276e+02 1.775e+02 2.154e+02 2.496e+02 3.947e+02, threshold=4.307e+02, percent-clipped=0.0 +2023-03-21 13:00:19,800 INFO [train.py:901] (1/2) Epoch 48, batch 350, loss[loss=0.1404, simple_loss=0.2178, pruned_loss=0.03144, over 7364.00 frames. ], tot_loss[loss=0.1266, simple_loss=0.2078, pruned_loss=0.02271, over 1194438.24 frames. ], batch size: 63, lr: 3.45e-03, grad_scale: 8.0 +2023-03-21 13:00:32,650 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133102.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:00:37,622 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 13:00:46,192 INFO [train.py:901] (1/2) Epoch 48, batch 400, loss[loss=0.125, simple_loss=0.2098, pruned_loss=0.02008, over 7354.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.208, pruned_loss=0.02264, over 1249992.01 frames. ], batch size: 63, lr: 3.45e-03, grad_scale: 8.0 +2023-03-21 13:01:04,085 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.252e+02 1.745e+02 2.126e+02 2.539e+02 4.672e+02, threshold=4.251e+02, percent-clipped=1.0 +2023-03-21 13:01:12,432 INFO [train.py:901] (1/2) Epoch 48, batch 450, loss[loss=0.0869, simple_loss=0.1535, pruned_loss=0.01014, over 5841.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2078, pruned_loss=0.02278, over 1291849.65 frames. ], batch size: 25, lr: 3.45e-03, grad_scale: 8.0 +2023-03-21 13:01:19,482 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 13:01:19,995 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 13:01:38,006 INFO [train.py:901] (1/2) Epoch 48, batch 500, loss[loss=0.1143, simple_loss=0.1965, pruned_loss=0.01601, over 7137.00 frames. ], tot_loss[loss=0.127, simple_loss=0.2082, pruned_loss=0.02291, over 1326176.77 frames. ], batch size: 41, lr: 3.45e-03, grad_scale: 8.0 +2023-03-21 13:01:53,346 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 13:01:54,838 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 13:01:55,398 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 13:01:57,519 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.202e+02 1.728e+02 2.008e+02 2.239e+02 3.859e+02, threshold=4.016e+02, percent-clipped=0.0 +2023-03-21 13:01:58,103 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 13:01:58,674 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2004, 4.3804, 4.1356, 4.4181, 3.9782, 4.3286, 4.6349, 4.6559], + device='cuda:1'), covar=tensor([0.0195, 0.0145, 0.0203, 0.0143, 0.0313, 0.0239, 0.0219, 0.0200], + device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0130, 0.0123, 0.0127, 0.0116, 0.0103, 0.0100, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 13:02:01,792 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8423, 2.7792, 3.9816, 3.7599, 3.9879, 3.9653, 4.0063, 3.5721], + device='cuda:1'), covar=tensor([0.0059, 0.0277, 0.0052, 0.0072, 0.0052, 0.0053, 0.0046, 0.0091], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0074, 0.0062, 0.0060, 0.0057, 0.0063, 0.0049, 0.0082], + device='cuda:1'), out_proj_covar=tensor([8.5669e-05, 1.4681e-04, 1.0784e-04, 1.0051e-04, 9.2275e-05, 1.0592e-04, + 9.1166e-05, 1.4699e-04], device='cuda:1') +2023-03-21 13:02:02,679 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 13:02:04,704 INFO [train.py:901] (1/2) Epoch 48, batch 550, loss[loss=0.1199, simple_loss=0.2124, pruned_loss=0.01368, over 7274.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2078, pruned_loss=0.0225, over 1352972.98 frames. ], batch size: 64, lr: 3.45e-03, grad_scale: 8.0 +2023-03-21 13:02:14,796 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 13:02:22,804 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-03-21 13:02:23,455 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 13:02:23,570 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133316.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:02:26,583 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 13:02:30,055 INFO [train.py:901] (1/2) Epoch 48, batch 600, loss[loss=0.1234, simple_loss=0.2114, pruned_loss=0.01769, over 7278.00 frames. ], tot_loss[loss=0.1258, simple_loss=0.2071, pruned_loss=0.02231, over 1372066.27 frames. ], batch size: 70, lr: 3.44e-03, grad_scale: 8.0 +2023-03-21 13:02:33,124 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 13:02:33,217 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9994, 3.7494, 3.6981, 3.6573, 3.7743, 3.5921, 3.9076, 3.5726], + device='cuda:1'), covar=tensor([0.0141, 0.0183, 0.0129, 0.0202, 0.0378, 0.0133, 0.0146, 0.0149], + device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0107, 0.0109, 0.0095, 0.0185, 0.0114, 0.0110, 0.0119], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 13:02:36,184 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4625, 2.5361, 2.7665, 2.4884, 2.6841, 2.3623, 2.2822, 2.2214], + device='cuda:1'), covar=tensor([0.0728, 0.0612, 0.0317, 0.0413, 0.0741, 0.0668, 0.0613, 0.0300], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0042, 0.0044, 0.0043, 0.0040, 0.0041, 0.0047, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 13:02:45,471 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0290, 2.5796, 3.4433, 2.9725, 3.2142, 2.9406, 2.5797, 3.0983], + device='cuda:1'), covar=tensor([0.1237, 0.0821, 0.0646, 0.1132, 0.0592, 0.0866, 0.1651, 0.1230], + device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0073, 0.0055, 0.0054, 0.0053, 0.0052, 0.0072, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 13:02:48,947 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=133364.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:02:49,400 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 1.711e+02 1.911e+02 2.129e+02 5.937e+02, threshold=3.823e+02, percent-clipped=2.0 +2023-03-21 13:02:50,413 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 13:02:54,077 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133374.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 13:02:55,130 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133376.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:02:56,503 INFO [train.py:901] (1/2) Epoch 48, batch 650, loss[loss=0.1039, simple_loss=0.1781, pruned_loss=0.01489, over 6956.00 frames. ], tot_loss[loss=0.1258, simple_loss=0.2071, pruned_loss=0.02224, over 1385300.82 frames. ], batch size: 35, lr: 3.44e-03, grad_scale: 8.0 +2023-03-21 13:02:59,597 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 13:03:08,678 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133402.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:03:16,796 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 13:03:22,960 INFO [train.py:901] (1/2) Epoch 48, batch 700, loss[loss=0.1361, simple_loss=0.2123, pruned_loss=0.02998, over 7323.00 frames. ], tot_loss[loss=0.1253, simple_loss=0.2069, pruned_loss=0.02192, over 1399732.84 frames. ], batch size: 54, lr: 3.44e-03, grad_scale: 8.0 +2023-03-21 13:03:26,258 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 13:03:26,409 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133435.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 13:03:27,390 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133437.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:03:34,489 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=133450.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:03:41,877 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.129e+02 1.729e+02 1.961e+02 2.406e+02 3.569e+02, threshold=3.922e+02, percent-clipped=0.0 +2023-03-21 13:03:46,600 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133474.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:03:49,052 INFO [train.py:901] (1/2) Epoch 48, batch 750, loss[loss=0.1237, simple_loss=0.2101, pruned_loss=0.01868, over 7323.00 frames. ], tot_loss[loss=0.1254, simple_loss=0.2071, pruned_loss=0.02189, over 1409896.28 frames. ], batch size: 83, lr: 3.44e-03, grad_scale: 8.0 +2023-03-21 13:03:50,051 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 13:03:50,549 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 13:04:04,086 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 13:04:09,224 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 13:04:15,410 INFO [train.py:901] (1/2) Epoch 48, batch 800, loss[loss=0.1361, simple_loss=0.226, pruned_loss=0.02308, over 7274.00 frames. ], tot_loss[loss=0.126, simple_loss=0.2079, pruned_loss=0.02208, over 1418528.40 frames. ], batch size: 77, lr: 3.44e-03, grad_scale: 8.0 +2023-03-21 13:04:15,960 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 13:04:16,948 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 13:04:18,590 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133535.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:04:27,407 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 13:04:33,573 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.190e+02 1.718e+02 2.033e+02 2.465e+02 3.831e+02, threshold=4.067e+02, percent-clipped=0.0 +2023-03-21 13:04:40,712 INFO [train.py:901] (1/2) Epoch 48, batch 850, loss[loss=0.09995, simple_loss=0.1767, pruned_loss=0.01162, over 7202.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2085, pruned_loss=0.0224, over 1423139.83 frames. ], batch size: 39, lr: 3.44e-03, grad_scale: 8.0 +2023-03-21 13:04:41,920 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4522, 2.8998, 2.2286, 3.2267, 3.1001, 3.2110, 2.8878, 2.8523], + device='cuda:1'), covar=tensor([0.2276, 0.1143, 0.3690, 0.0687, 0.0443, 0.0313, 0.0591, 0.0434], + device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0231, 0.0243, 0.0255, 0.0203, 0.0206, 0.0219, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:04:46,357 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 13:04:46,847 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 13:04:52,844 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 13:04:56,332 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 13:04:58,460 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8688, 1.7507, 2.0882, 2.4137, 2.1937, 2.2530, 2.1336, 2.3729], + device='cuda:1'), covar=tensor([0.2624, 0.5685, 0.2470, 0.1314, 0.1478, 0.2587, 0.2105, 0.2056], + device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0086, 0.0080, 0.0069, 0.0070, 0.0070, 0.0110, 0.0071], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 13:05:04,607 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9283, 3.1914, 3.9435, 3.8846, 3.9835, 4.0284, 4.0529, 3.8228], + device='cuda:1'), covar=tensor([0.0038, 0.0143, 0.0038, 0.0039, 0.0036, 0.0034, 0.0039, 0.0056], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0074, 0.0062, 0.0060, 0.0056, 0.0062, 0.0049, 0.0081], + device='cuda:1'), out_proj_covar=tensor([8.4869e-05, 1.4659e-04, 1.0762e-04, 9.9954e-05, 9.1868e-05, 1.0526e-04, + 9.0349e-05, 1.4594e-04], device='cuda:1') +2023-03-21 13:05:07,527 INFO [train.py:901] (1/2) Epoch 48, batch 900, loss[loss=0.1224, simple_loss=0.2025, pruned_loss=0.02116, over 7326.00 frames. ], tot_loss[loss=0.1266, simple_loss=0.2079, pruned_loss=0.02263, over 1427791.01 frames. ], batch size: 49, lr: 3.44e-03, grad_scale: 8.0 +2023-03-21 13:05:19,389 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133652.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:05:23,875 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1695, 4.6482, 4.7430, 4.6218, 4.6296, 4.1613, 4.7245, 4.5804], + device='cuda:1'), covar=tensor([0.0492, 0.0431, 0.0405, 0.0577, 0.0350, 0.0504, 0.0388, 0.0482], + device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0274, 0.0214, 0.0212, 0.0163, 0.0243, 0.0221, 0.0153], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 13:05:25,799 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.155e+02 1.711e+02 1.987e+02 2.449e+02 8.050e+02, threshold=3.974e+02, percent-clipped=6.0 +2023-03-21 13:05:29,029 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8899, 3.8154, 2.9285, 3.5067, 2.8815, 2.1861, 1.8188, 3.8287], + device='cuda:1'), covar=tensor([0.0057, 0.0062, 0.0202, 0.0084, 0.0229, 0.0726, 0.0804, 0.0076], + device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0096, 0.0119, 0.0101, 0.0137, 0.0137, 0.0133, 0.0109], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 13:05:32,846 INFO [train.py:901] (1/2) Epoch 48, batch 950, loss[loss=0.1147, simple_loss=0.2069, pruned_loss=0.01128, over 7267.00 frames. ], tot_loss[loss=0.1254, simple_loss=0.2064, pruned_loss=0.02219, over 1428089.54 frames. ], batch size: 64, lr: 3.44e-03, grad_scale: 8.0 +2023-03-21 13:05:35,460 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 13:05:51,697 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133713.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:05:59,173 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 13:05:59,683 INFO [train.py:901] (1/2) Epoch 48, batch 1000, loss[loss=0.13, simple_loss=0.2125, pruned_loss=0.02372, over 7297.00 frames. ], tot_loss[loss=0.125, simple_loss=0.2064, pruned_loss=0.02181, over 1430570.91 frames. ], batch size: 68, lr: 3.44e-03, grad_scale: 8.0 +2023-03-21 13:06:00,285 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133730.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 13:06:01,337 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133732.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:06:17,970 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.525e+02 1.854e+02 2.105e+02 4.157e+02, threshold=3.709e+02, percent-clipped=1.0 +2023-03-21 13:06:18,509 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 13:06:25,678 INFO [train.py:901] (1/2) Epoch 48, batch 1050, loss[loss=0.1128, simple_loss=0.189, pruned_loss=0.01831, over 7156.00 frames. ], tot_loss[loss=0.125, simple_loss=0.2064, pruned_loss=0.02178, over 1434567.90 frames. ], batch size: 41, lr: 3.44e-03, grad_scale: 8.0 +2023-03-21 13:06:39,946 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 +2023-03-21 13:06:41,706 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 13:06:46,188 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 13:06:48,846 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0285, 3.3717, 2.8631, 3.1116, 3.1049, 2.6672, 2.9240, 3.0510], + device='cuda:1'), covar=tensor([0.0878, 0.0522, 0.1173, 0.0960, 0.0781, 0.0910, 0.1448, 0.1312], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0062, 0.0070, 0.0062, 0.0059, 0.0066, 0.0059, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:06:51,711 INFO [train.py:901] (1/2) Epoch 48, batch 1100, loss[loss=0.1267, simple_loss=0.2109, pruned_loss=0.02131, over 7122.00 frames. ], tot_loss[loss=0.125, simple_loss=0.2063, pruned_loss=0.02189, over 1437385.92 frames. ], batch size: 98, lr: 3.44e-03, grad_scale: 8.0 +2023-03-21 13:06:52,302 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133830.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:07:11,023 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 +2023-03-21 13:07:11,124 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.249e+02 1.881e+02 2.184e+02 2.691e+02 4.315e+02, threshold=4.367e+02, percent-clipped=4.0 +2023-03-21 13:07:13,768 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.1688, 4.2936, 4.0777, 4.2419, 3.8151, 4.2228, 4.5353, 4.5589], + device='cuda:1'), covar=tensor([0.0176, 0.0134, 0.0210, 0.0145, 0.0393, 0.0274, 0.0241, 0.0181], + device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0130, 0.0125, 0.0127, 0.0116, 0.0104, 0.0101, 0.0104], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 13:07:14,642 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 13:07:15,139 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 13:07:18,219 INFO [train.py:901] (1/2) Epoch 48, batch 1150, loss[loss=0.1164, simple_loss=0.2, pruned_loss=0.01638, over 7245.00 frames. ], tot_loss[loss=0.1248, simple_loss=0.2059, pruned_loss=0.02185, over 1435333.01 frames. ], batch size: 45, lr: 3.44e-03, grad_scale: 16.0 +2023-03-21 13:07:26,706 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 13:07:27,695 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 13:07:40,498 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8955, 4.3501, 4.3837, 4.3315, 4.4085, 4.0224, 4.3989, 4.3545], + device='cuda:1'), covar=tensor([0.0524, 0.0462, 0.0452, 0.0598, 0.0274, 0.0468, 0.0408, 0.0366], + device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0274, 0.0214, 0.0212, 0.0162, 0.0242, 0.0222, 0.0154], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 13:07:43,414 INFO [train.py:901] (1/2) Epoch 48, batch 1200, loss[loss=0.136, simple_loss=0.2213, pruned_loss=0.02536, over 7346.00 frames. ], tot_loss[loss=0.1251, simple_loss=0.2064, pruned_loss=0.02189, over 1434537.62 frames. ], batch size: 61, lr: 3.44e-03, grad_scale: 16.0 +2023-03-21 13:07:51,526 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 +2023-03-21 13:08:02,812 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 13:08:03,295 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.228e+02 1.806e+02 2.192e+02 2.530e+02 3.927e+02, threshold=4.384e+02, percent-clipped=0.0 +2023-03-21 13:08:04,163 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 +2023-03-21 13:08:09,881 INFO [train.py:901] (1/2) Epoch 48, batch 1250, loss[loss=0.1228, simple_loss=0.2043, pruned_loss=0.0206, over 7288.00 frames. ], tot_loss[loss=0.1256, simple_loss=0.2068, pruned_loss=0.02224, over 1436183.63 frames. ], batch size: 66, lr: 3.44e-03, grad_scale: 8.0 +2023-03-21 13:08:19,042 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133997.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:08:24,919 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134008.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:08:26,377 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 13:08:30,981 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 13:08:32,012 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 13:08:36,306 INFO [train.py:901] (1/2) Epoch 48, batch 1300, loss[loss=0.1471, simple_loss=0.2353, pruned_loss=0.0295, over 7213.00 frames. ], tot_loss[loss=0.1265, simple_loss=0.208, pruned_loss=0.02244, over 1439944.81 frames. ], batch size: 99, lr: 3.44e-03, grad_scale: 8.0 +2023-03-21 13:08:36,923 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134030.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 13:08:37,991 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134032.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:08:51,821 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134058.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:08:55,688 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.358e+02 1.728e+02 2.012e+02 2.304e+02 3.451e+02, threshold=4.025e+02, percent-clipped=0.0 +2023-03-21 13:08:56,226 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 13:08:58,280 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 13:09:01,799 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 13:09:01,853 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=134078.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 13:09:02,247 INFO [train.py:901] (1/2) Epoch 48, batch 1350, loss[loss=0.1452, simple_loss=0.2272, pruned_loss=0.03159, over 7374.00 frames. ], tot_loss[loss=0.127, simple_loss=0.2087, pruned_loss=0.02265, over 1442548.58 frames. ], batch size: 65, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:09:02,817 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=134080.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:09:11,556 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 13:09:28,303 INFO [train.py:901] (1/2) Epoch 48, batch 1400, loss[loss=0.1204, simple_loss=0.2009, pruned_loss=0.01999, over 7305.00 frames. ], tot_loss[loss=0.1272, simple_loss=0.2086, pruned_loss=0.02293, over 1441538.36 frames. ], batch size: 83, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:09:29,568 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134130.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:09:44,782 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 13:09:47,824 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.303e+02 1.834e+02 2.048e+02 2.397e+02 3.522e+02, threshold=4.097e+02, percent-clipped=0.0 +2023-03-21 13:09:54,047 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=134178.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:09:54,483 INFO [train.py:901] (1/2) Epoch 48, batch 1450, loss[loss=0.1245, simple_loss=0.2086, pruned_loss=0.02021, over 7255.00 frames. ], tot_loss[loss=0.1273, simple_loss=0.2088, pruned_loss=0.02296, over 1442610.42 frames. ], batch size: 64, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:10:09,694 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 13:10:21,519 INFO [train.py:901] (1/2) Epoch 48, batch 1500, loss[loss=0.1257, simple_loss=0.2104, pruned_loss=0.02048, over 7275.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2084, pruned_loss=0.02251, over 1443015.35 frames. ], batch size: 77, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:10:27,080 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 13:10:33,456 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 +2023-03-21 13:10:34,946 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.10 vs. limit=5.0 +2023-03-21 13:10:40,282 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.207e+02 1.705e+02 1.935e+02 2.252e+02 3.749e+02, threshold=3.870e+02, percent-clipped=0.0 +2023-03-21 13:10:46,852 INFO [train.py:901] (1/2) Epoch 48, batch 1550, loss[loss=0.1393, simple_loss=0.2168, pruned_loss=0.03086, over 7353.00 frames. ], tot_loss[loss=0.1261, simple_loss=0.2079, pruned_loss=0.02216, over 1443555.96 frames. ], batch size: 51, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:10:52,036 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 13:11:00,293 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134304.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:11:03,045 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134308.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:11:03,589 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134309.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:11:13,634 INFO [train.py:901] (1/2) Epoch 48, batch 1600, loss[loss=0.1288, simple_loss=0.2152, pruned_loss=0.02119, over 7342.00 frames. ], tot_loss[loss=0.1266, simple_loss=0.2083, pruned_loss=0.02242, over 1443079.54 frames. ], batch size: 54, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:11:23,820 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 13:11:24,831 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 13:11:25,935 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134353.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:11:27,431 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 13:11:27,465 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=134356.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:11:28,565 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3261, 2.4870, 2.6150, 2.3884, 2.6301, 2.5706, 2.3639, 2.0568], + device='cuda:1'), covar=tensor([0.0604, 0.0420, 0.0437, 0.0343, 0.0763, 0.0696, 0.0380, 0.0455], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0042, 0.0043, 0.0043, 0.0041, 0.0041, 0.0047, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 13:11:32,110 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134365.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:11:32,447 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.266e+02 1.662e+02 1.962e+02 2.325e+02 5.415e+02, threshold=3.925e+02, percent-clipped=1.0 +2023-03-21 13:11:35,252 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134370.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:11:37,224 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 13:11:39,765 INFO [train.py:901] (1/2) Epoch 48, batch 1650, loss[loss=0.1314, simple_loss=0.2152, pruned_loss=0.02381, over 7342.00 frames. ], tot_loss[loss=0.1269, simple_loss=0.2085, pruned_loss=0.02266, over 1441483.35 frames. ], batch size: 63, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:11:41,863 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 13:11:46,594 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0025, 3.2615, 2.6926, 3.3805, 3.0079, 2.8846, 3.1057, 3.0560], + device='cuda:1'), covar=tensor([0.0773, 0.0499, 0.0973, 0.0601, 0.1298, 0.0644, 0.1007, 0.0980], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0063, 0.0070, 0.0062, 0.0059, 0.0066, 0.0059, 0.0057], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:11:50,668 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 13:12:06,334 INFO [train.py:901] (1/2) Epoch 48, batch 1700, loss[loss=0.1235, simple_loss=0.2072, pruned_loss=0.01991, over 7362.00 frames. ], tot_loss[loss=0.1269, simple_loss=0.2083, pruned_loss=0.02272, over 1442787.02 frames. ], batch size: 54, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:12:07,883 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 13:12:11,946 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 13:12:23,048 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 13:12:25,575 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.287e+02 1.707e+02 1.940e+02 2.259e+02 3.650e+02, threshold=3.879e+02, percent-clipped=0.0 +2023-03-21 13:12:28,637 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-03-21 13:12:31,971 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4802, 2.4467, 2.8547, 2.5087, 2.8157, 2.7014, 2.3794, 2.2153], + device='cuda:1'), covar=tensor([0.0405, 0.0695, 0.0309, 0.0474, 0.0454, 0.0439, 0.0513, 0.0424], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0042, 0.0043, 0.0043, 0.0040, 0.0041, 0.0047, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 13:12:32,357 INFO [train.py:901] (1/2) Epoch 48, batch 1750, loss[loss=0.1147, simple_loss=0.2002, pruned_loss=0.01461, over 7337.00 frames. ], tot_loss[loss=0.1269, simple_loss=0.2084, pruned_loss=0.02274, over 1442886.11 frames. ], batch size: 54, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:12:39,916 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-03-21 13:12:48,226 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 13:12:48,774 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 13:12:58,287 INFO [train.py:901] (1/2) Epoch 48, batch 1800, loss[loss=0.1357, simple_loss=0.2174, pruned_loss=0.02701, over 7280.00 frames. ], tot_loss[loss=0.1271, simple_loss=0.2087, pruned_loss=0.02281, over 1443056.35 frames. ], batch size: 86, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:13:09,660 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 13:13:13,920 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-03-21 13:13:17,789 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.203e+02 1.829e+02 2.119e+02 2.568e+02 4.020e+02, threshold=4.237e+02, percent-clipped=1.0 +2023-03-21 13:13:24,629 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 13:13:25,123 INFO [train.py:901] (1/2) Epoch 48, batch 1850, loss[loss=0.1207, simple_loss=0.209, pruned_loss=0.01618, over 7288.00 frames. ], tot_loss[loss=0.1266, simple_loss=0.2083, pruned_loss=0.02246, over 1440235.84 frames. ], batch size: 77, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:13:34,360 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 13:13:51,704 INFO [train.py:901] (1/2) Epoch 48, batch 1900, loss[loss=0.1302, simple_loss=0.2141, pruned_loss=0.02312, over 7283.00 frames. ], tot_loss[loss=0.127, simple_loss=0.2088, pruned_loss=0.02255, over 1441111.94 frames. ], batch size: 68, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:13:51,706 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 13:13:52,874 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8085, 3.1115, 2.7125, 3.0230, 2.9604, 2.7274, 3.0778, 2.7892], + device='cuda:1'), covar=tensor([0.0710, 0.0891, 0.0993, 0.0723, 0.0979, 0.0530, 0.0544, 0.1283], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0062, 0.0070, 0.0062, 0.0058, 0.0066, 0.0059, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:14:04,117 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134653.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:14:08,326 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134660.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:14:10,853 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134665.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:14:11,274 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.369e+02 1.733e+02 1.999e+02 2.386e+02 3.499e+02, threshold=3.997e+02, percent-clipped=0.0 +2023-03-21 13:14:12,896 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0170, 4.6085, 4.4014, 4.9891, 4.7865, 4.9338, 4.3795, 4.5760], + device='cuda:1'), covar=tensor([0.0863, 0.2436, 0.2070, 0.0979, 0.0920, 0.1176, 0.0900, 0.1090], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0406, 0.0306, 0.0320, 0.0237, 0.0383, 0.0240, 0.0291], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 13:14:16,902 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 13:14:17,037 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.3953, 1.6765, 1.3733, 1.5357, 1.6248, 1.6050, 1.4646, 1.1939], + device='cuda:1'), covar=tensor([0.0234, 0.0176, 0.0304, 0.0217, 0.0162, 0.0139, 0.0186, 0.0234], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0042, 0.0040, 0.0042, 0.0039, 0.0039, 0.0041, 0.0050], + device='cuda:1'), out_proj_covar=tensor([4.7710e-05, 4.6151e-05, 4.4531e-05, 4.5875e-05, 4.3481e-05, 4.2793e-05, + 4.5601e-05, 5.4966e-05], device='cuda:1') +2023-03-21 13:14:17,901 INFO [train.py:901] (1/2) Epoch 48, batch 1950, loss[loss=0.1289, simple_loss=0.2156, pruned_loss=0.02112, over 7251.00 frames. ], tot_loss[loss=0.1266, simple_loss=0.2085, pruned_loss=0.02237, over 1441107.92 frames. ], batch size: 55, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:14:18,354 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 +2023-03-21 13:14:27,780 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9074, 2.6770, 2.6272, 3.9638, 2.1109, 3.6947, 1.5193, 3.2045], + device='cuda:1'), covar=tensor([0.0185, 0.1382, 0.2033, 0.0196, 0.4303, 0.0286, 0.1405, 0.0468], + device='cuda:1'), in_proj_covar=tensor([0.0211, 0.0244, 0.0257, 0.0213, 0.0250, 0.0222, 0.0220, 0.0232], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:14:28,154 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 13:14:29,199 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=134701.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:14:32,663 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 13:14:33,122 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 13:14:43,883 INFO [train.py:901] (1/2) Epoch 48, batch 2000, loss[loss=0.1291, simple_loss=0.2101, pruned_loss=0.02402, over 7293.00 frames. ], tot_loss[loss=0.1266, simple_loss=0.2084, pruned_loss=0.02244, over 1441792.03 frames. ], batch size: 86, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:14:49,643 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 13:15:00,385 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 13:15:03,340 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.689e+02 2.059e+02 2.364e+02 4.863e+02, threshold=4.119e+02, percent-clipped=2.0 +2023-03-21 13:15:04,585 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6583, 1.9657, 1.7049, 1.8064, 1.8890, 1.8987, 1.7785, 1.4571], + device='cuda:1'), covar=tensor([0.0202, 0.0203, 0.0286, 0.0280, 0.0177, 0.0152, 0.0204, 0.0230], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0041, 0.0040, 0.0042, 0.0039, 0.0038, 0.0041, 0.0050], + device='cuda:1'), out_proj_covar=tensor([4.7589e-05, 4.5873e-05, 4.4553e-05, 4.5812e-05, 4.3331e-05, 4.2446e-05, + 4.5488e-05, 5.4745e-05], device='cuda:1') +2023-03-21 13:15:07,954 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 13:15:09,988 INFO [train.py:901] (1/2) Epoch 48, batch 2050, loss[loss=0.1215, simple_loss=0.205, pruned_loss=0.01897, over 7250.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2084, pruned_loss=0.02219, over 1441234.10 frames. ], batch size: 70, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:15:30,818 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3972, 2.0474, 2.3842, 3.4702, 1.9589, 3.2506, 1.3433, 3.0852], + device='cuda:1'), covar=tensor([0.0257, 0.1821, 0.1835, 0.0238, 0.4111, 0.0316, 0.1406, 0.0529], + device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0244, 0.0255, 0.0212, 0.0249, 0.0221, 0.0219, 0.0232], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:15:36,856 INFO [train.py:901] (1/2) Epoch 48, batch 2100, loss[loss=0.1196, simple_loss=0.2022, pruned_loss=0.01848, over 7312.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2086, pruned_loss=0.02243, over 1443471.07 frames. ], batch size: 83, lr: 3.43e-03, grad_scale: 8.0 +2023-03-21 13:15:42,004 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 13:15:45,083 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 13:15:53,991 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134862.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:15:55,917 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.812e+02 2.170e+02 2.535e+02 4.024e+02, threshold=4.341e+02, percent-clipped=0.0 +2023-03-21 13:16:02,489 INFO [train.py:901] (1/2) Epoch 48, batch 2150, loss[loss=0.1107, simple_loss=0.1955, pruned_loss=0.01296, over 7278.00 frames. ], tot_loss[loss=0.1265, simple_loss=0.2082, pruned_loss=0.02233, over 1444114.39 frames. ], batch size: 47, lr: 3.42e-03, grad_scale: 8.0 +2023-03-21 13:16:26,352 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134923.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:16:29,185 INFO [train.py:901] (1/2) Epoch 48, batch 2200, loss[loss=0.1375, simple_loss=0.2198, pruned_loss=0.02762, over 7275.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2081, pruned_loss=0.02233, over 1443908.71 frames. ], batch size: 66, lr: 3.42e-03, grad_scale: 8.0 +2023-03-21 13:16:32,307 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 13:16:45,164 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134960.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:16:47,629 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134965.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:16:48,035 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 1.726e+02 2.022e+02 2.434e+02 4.958e+02, threshold=4.044e+02, percent-clipped=3.0 +2023-03-21 13:16:55,315 INFO [train.py:901] (1/2) Epoch 48, batch 2250, loss[loss=0.1425, simple_loss=0.2266, pruned_loss=0.02921, over 7288.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2086, pruned_loss=0.02247, over 1443110.38 frames. ], batch size: 66, lr: 3.42e-03, grad_scale: 8.0 +2023-03-21 13:17:05,221 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7550, 3.0054, 3.6839, 3.7264, 3.7933, 3.7270, 3.7324, 3.6961], + device='cuda:1'), covar=tensor([0.0033, 0.0138, 0.0036, 0.0035, 0.0034, 0.0034, 0.0062, 0.0048], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0074, 0.0062, 0.0059, 0.0057, 0.0063, 0.0049, 0.0082], + device='cuda:1'), out_proj_covar=tensor([8.5761e-05, 1.4568e-04, 1.0701e-04, 9.8062e-05, 9.2784e-05, 1.0611e-04, + 9.0356e-05, 1.4682e-04], device='cuda:1') +2023-03-21 13:17:05,267 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5719, 1.8011, 1.5563, 1.6888, 1.7902, 1.8084, 1.7263, 1.4241], + device='cuda:1'), covar=tensor([0.0202, 0.0223, 0.0326, 0.0267, 0.0171, 0.0161, 0.0323, 0.0209], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0041, 0.0039, 0.0041, 0.0039, 0.0038, 0.0040, 0.0049], + device='cuda:1'), out_proj_covar=tensor([4.7063e-05, 4.5281e-05, 4.4059e-05, 4.4961e-05, 4.2547e-05, 4.2156e-05, + 4.4906e-05, 5.4209e-05], device='cuda:1') +2023-03-21 13:17:06,650 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 13:17:07,447 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 13:17:11,149 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=135008.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:17:13,647 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=135013.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:17:20,718 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 13:17:21,761 INFO [train.py:901] (1/2) Epoch 48, batch 2300, loss[loss=0.1121, simple_loss=0.199, pruned_loss=0.01259, over 7279.00 frames. ], tot_loss[loss=0.1269, simple_loss=0.2085, pruned_loss=0.02268, over 1442825.71 frames. ], batch size: 68, lr: 3.42e-03, grad_scale: 8.0 +2023-03-21 13:17:23,509 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9539, 3.6782, 3.5998, 3.6328, 3.6938, 3.5023, 3.8254, 3.4891], + device='cuda:1'), covar=tensor([0.0132, 0.0187, 0.0129, 0.0206, 0.0381, 0.0124, 0.0154, 0.0167], + device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0108, 0.0109, 0.0095, 0.0186, 0.0114, 0.0111, 0.0119], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 13:17:41,371 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.741e+02 2.077e+02 2.361e+02 4.770e+02, threshold=4.154e+02, percent-clipped=2.0 +2023-03-21 13:17:41,547 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7800, 2.1916, 1.7993, 2.1173, 2.1600, 1.8729, 1.9231, 1.5995], + device='cuda:1'), covar=tensor([0.0176, 0.0173, 0.0267, 0.0182, 0.0115, 0.0188, 0.0250, 0.0272], + device='cuda:1'), in_proj_covar=tensor([0.0042, 0.0041, 0.0039, 0.0041, 0.0039, 0.0038, 0.0040, 0.0050], + device='cuda:1'), out_proj_covar=tensor([4.7253e-05, 4.5576e-05, 4.4231e-05, 4.5098e-05, 4.2649e-05, 4.2423e-05, + 4.5075e-05, 5.4374e-05], device='cuda:1') +2023-03-21 13:17:43,279 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 +2023-03-21 13:17:47,946 INFO [train.py:901] (1/2) Epoch 48, batch 2350, loss[loss=0.1448, simple_loss=0.2257, pruned_loss=0.03197, over 7310.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2081, pruned_loss=0.02265, over 1441192.59 frames. ], batch size: 49, lr: 3.42e-03, grad_scale: 8.0 +2023-03-21 13:17:58,077 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-21 13:18:08,364 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 13:18:13,927 INFO [train.py:901] (1/2) Epoch 48, batch 2400, loss[loss=0.1277, simple_loss=0.2178, pruned_loss=0.01881, over 7212.00 frames. ], tot_loss[loss=0.127, simple_loss=0.2084, pruned_loss=0.02281, over 1440660.37 frames. ], batch size: 93, lr: 3.42e-03, grad_scale: 8.0 +2023-03-21 13:18:14,979 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 13:18:26,351 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 13:18:29,409 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 13:18:33,397 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.237e+02 1.657e+02 2.114e+02 2.439e+02 4.545e+02, threshold=4.228e+02, percent-clipped=1.0 +2023-03-21 13:18:36,807 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135171.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:18:40,684 INFO [train.py:901] (1/2) Epoch 48, batch 2450, loss[loss=0.1428, simple_loss=0.2196, pruned_loss=0.03297, over 7326.00 frames. ], tot_loss[loss=0.1263, simple_loss=0.2078, pruned_loss=0.0224, over 1441613.33 frames. ], batch size: 54, lr: 3.42e-03, grad_scale: 8.0 +2023-03-21 13:18:43,317 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4983, 4.2557, 3.6388, 3.9930, 3.4326, 2.5332, 1.9822, 4.4550], + device='cuda:1'), covar=tensor([0.0048, 0.0058, 0.0140, 0.0072, 0.0151, 0.0604, 0.0702, 0.0048], + device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0094, 0.0117, 0.0100, 0.0134, 0.0134, 0.0131, 0.0106], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 13:18:56,415 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 13:19:01,121 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135218.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:19:06,641 INFO [train.py:901] (1/2) Epoch 48, batch 2500, loss[loss=0.1137, simple_loss=0.1941, pruned_loss=0.01666, over 7351.00 frames. ], tot_loss[loss=0.1258, simple_loss=0.2075, pruned_loss=0.02201, over 1441481.30 frames. ], batch size: 44, lr: 3.42e-03, grad_scale: 8.0 +2023-03-21 13:19:08,343 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135232.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:19:24,338 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 13:19:26,855 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.741e+02 2.067e+02 2.417e+02 6.629e+02, threshold=4.134e+02, percent-clipped=1.0 +2023-03-21 13:19:33,505 INFO [train.py:901] (1/2) Epoch 48, batch 2550, loss[loss=0.1284, simple_loss=0.2022, pruned_loss=0.02734, over 7363.00 frames. ], tot_loss[loss=0.1258, simple_loss=0.2072, pruned_loss=0.0222, over 1441156.39 frames. ], batch size: 63, lr: 3.42e-03, grad_scale: 8.0 +2023-03-21 13:19:44,118 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3273, 2.5910, 2.8015, 2.4544, 2.5367, 2.5121, 2.3483, 2.1440], + device='cuda:1'), covar=tensor([0.0576, 0.0397, 0.0242, 0.0330, 0.0692, 0.0454, 0.0420, 0.0347], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0043, 0.0044, 0.0043, 0.0041, 0.0041, 0.0048, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 13:19:50,280 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9516, 3.2018, 2.8552, 3.0454, 3.1373, 2.8124, 3.1167, 3.0433], + device='cuda:1'), covar=tensor([0.0624, 0.0601, 0.0836, 0.1429, 0.0916, 0.0576, 0.0543, 0.1092], + device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0063, 0.0070, 0.0063, 0.0059, 0.0067, 0.0060, 0.0057], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:19:58,865 INFO [train.py:901] (1/2) Epoch 48, batch 2600, loss[loss=0.1188, simple_loss=0.2081, pruned_loss=0.01481, over 7245.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.2071, pruned_loss=0.0221, over 1441424.88 frames. ], batch size: 55, lr: 3.42e-03, grad_scale: 8.0 +2023-03-21 13:20:04,043 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5542, 4.9550, 5.0859, 5.0043, 4.8911, 4.4861, 5.0649, 4.8120], + device='cuda:1'), covar=tensor([0.0443, 0.0363, 0.0317, 0.0400, 0.0363, 0.0379, 0.0329, 0.0497], + device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0270, 0.0210, 0.0209, 0.0160, 0.0238, 0.0219, 0.0151], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 13:20:10,923 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 +2023-03-21 13:20:18,009 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.707e+02 2.022e+02 2.362e+02 4.542e+02, threshold=4.043e+02, percent-clipped=1.0 +2023-03-21 13:20:24,536 INFO [train.py:901] (1/2) Epoch 48, batch 2650, loss[loss=0.1065, simple_loss=0.1865, pruned_loss=0.01325, over 7160.00 frames. ], tot_loss[loss=0.1255, simple_loss=0.2067, pruned_loss=0.02218, over 1442212.19 frames. ], batch size: 39, lr: 3.42e-03, grad_scale: 8.0 +2023-03-21 13:20:36,993 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6926, 3.0425, 2.7060, 2.8068, 2.9486, 2.5159, 3.0286, 2.8779], + device='cuda:1'), covar=tensor([0.0909, 0.0720, 0.0810, 0.1565, 0.1285, 0.0993, 0.1179, 0.0947], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0062, 0.0070, 0.0062, 0.0059, 0.0067, 0.0059, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:20:49,776 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 +2023-03-21 13:20:50,014 INFO [train.py:901] (1/2) Epoch 48, batch 2700, loss[loss=0.1241, simple_loss=0.2094, pruned_loss=0.01937, over 7327.00 frames. ], tot_loss[loss=0.1263, simple_loss=0.2076, pruned_loss=0.02251, over 1441506.96 frames. ], batch size: 75, lr: 3.42e-03, grad_scale: 8.0 +2023-03-21 13:21:00,801 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 +2023-03-21 13:21:05,695 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-03-21 13:21:08,366 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.481e+02 1.782e+02 2.093e+02 2.517e+02 6.149e+02, threshold=4.186e+02, percent-clipped=2.0 +2023-03-21 13:21:14,688 INFO [train.py:901] (1/2) Epoch 48, batch 2750, loss[loss=0.121, simple_loss=0.1927, pruned_loss=0.02467, over 7318.00 frames. ], tot_loss[loss=0.127, simple_loss=0.2082, pruned_loss=0.02287, over 1442723.72 frames. ], batch size: 49, lr: 3.42e-03, grad_scale: 8.0 +2023-03-21 13:21:34,357 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135518.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:21:38,625 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135527.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:21:39,561 INFO [train.py:901] (1/2) Epoch 48, batch 2800, loss[loss=0.1215, simple_loss=0.2035, pruned_loss=0.01971, over 7256.00 frames. ], tot_loss[loss=0.127, simple_loss=0.2084, pruned_loss=0.0228, over 1440794.58 frames. ], batch size: 47, lr: 3.42e-03, grad_scale: 8.0 +2023-03-21 13:22:02,865 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 13:22:02,930 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 13:22:02,956 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 13:22:02,972 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 13:22:03,460 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 13:22:03,477 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 13:22:03,488 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 13:22:03,584 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 13:22:03,659 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 13:22:03,673 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 13:22:03,691 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 13:22:03,849 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 13:22:03,856 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 13:22:03,863 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 13:22:03,999 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 13:22:04,001 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 13:22:04,059 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 13:22:08,210 INFO [train.py:901] (1/2) Epoch 49, batch 0, loss[loss=0.1252, simple_loss=0.2159, pruned_loss=0.01729, over 7283.00 frames. ], tot_loss[loss=0.1252, simple_loss=0.2159, pruned_loss=0.01729, over 7283.00 frames. ], batch size: 70, lr: 3.38e-03, grad_scale: 8.0 +2023-03-21 13:22:08,211 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 13:22:14,434 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.9404, 3.5924, 4.0617, 4.1798, 4.1700, 3.9643, 4.5091, 4.1389], + device='cuda:1'), covar=tensor([0.0037, 0.0111, 0.0036, 0.0029, 0.0030, 0.0037, 0.0014, 0.0042], + device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0074, 0.0062, 0.0060, 0.0057, 0.0063, 0.0049, 0.0082], + device='cuda:1'), out_proj_covar=tensor([8.5679e-05, 1.4646e-04, 1.0733e-04, 9.8756e-05, 9.2221e-05, 1.0694e-04, + 9.0068e-05, 1.4766e-04], device='cuda:1') +2023-03-21 13:22:21,341 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.1481, 4.7262, 4.4606, 5.1427, 4.8743, 5.1491, 4.5541, 4.9600], + device='cuda:1'), covar=tensor([0.0575, 0.1839, 0.1700, 0.0994, 0.0635, 0.0754, 0.0541, 0.0647], + device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0405, 0.0305, 0.0318, 0.0238, 0.0382, 0.0238, 0.0287], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 13:22:26,004 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.5140, 2.5411, 2.5719, 3.6932, 2.0780, 3.4218, 1.5282, 3.3711], + device='cuda:1'), covar=tensor([0.0165, 0.1517, 0.1886, 0.0210, 0.3811, 0.0278, 0.1349, 0.0409], + device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0241, 0.0253, 0.0212, 0.0246, 0.0218, 0.0216, 0.0228], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:22:32,974 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9030, 3.1631, 2.8241, 3.0306, 3.0794, 2.6940, 3.0221, 3.1191], + device='cuda:1'), covar=tensor([0.0570, 0.0401, 0.0641, 0.0885, 0.0618, 0.0688, 0.0752, 0.0531], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0062, 0.0070, 0.0062, 0.0058, 0.0066, 0.0059, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:22:33,727 INFO [train.py:935] (1/2) Epoch 49, validation: loss=0.1665, simple_loss=0.2593, pruned_loss=0.03685, over 1622729.00 frames. +2023-03-21 13:22:33,727 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 13:22:40,313 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.219e+02 1.655e+02 2.021e+02 2.392e+02 5.342e+02, threshold=4.041e+02, percent-clipped=1.0 +2023-03-21 13:22:40,341 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 13:22:40,386 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=135566.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:22:51,596 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 13:22:58,900 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 13:23:00,815 INFO [train.py:901] (1/2) Epoch 49, batch 50, loss[loss=0.1318, simple_loss=0.2151, pruned_loss=0.02422, over 7270.00 frames. ], tot_loss[loss=0.1231, simple_loss=0.2066, pruned_loss=0.01984, over 326606.47 frames. ], batch size: 64, lr: 3.38e-03, grad_scale: 8.0 +2023-03-21 13:23:01,891 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 13:23:04,929 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 13:23:10,729 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 +2023-03-21 13:23:14,681 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135630.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:23:21,969 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 13:23:22,507 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 13:23:26,004 INFO [train.py:901] (1/2) Epoch 49, batch 100, loss[loss=0.1205, simple_loss=0.2015, pruned_loss=0.01975, over 7285.00 frames. ], tot_loss[loss=0.1239, simple_loss=0.2064, pruned_loss=0.02072, over 573956.09 frames. ], batch size: 47, lr: 3.38e-03, grad_scale: 8.0 +2023-03-21 13:23:32,727 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.252e+02 1.833e+02 2.096e+02 2.422e+02 4.356e+02, threshold=4.193e+02, percent-clipped=1.0 +2023-03-21 13:23:47,043 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135691.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:23:52,949 INFO [train.py:901] (1/2) Epoch 49, batch 150, loss[loss=0.1211, simple_loss=0.2121, pruned_loss=0.01501, over 7284.00 frames. ], tot_loss[loss=0.1246, simple_loss=0.2066, pruned_loss=0.0213, over 766553.12 frames. ], batch size: 57, lr: 3.38e-03, grad_scale: 8.0 +2023-03-21 13:23:55,242 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-03-21 13:24:18,776 INFO [train.py:901] (1/2) Epoch 49, batch 200, loss[loss=0.1184, simple_loss=0.1949, pruned_loss=0.02097, over 7339.00 frames. ], tot_loss[loss=0.1245, simple_loss=0.2059, pruned_loss=0.02158, over 913116.59 frames. ], batch size: 44, lr: 3.38e-03, grad_scale: 8.0 +2023-03-21 13:24:24,356 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 13:24:25,363 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.280e+02 1.715e+02 2.006e+02 2.248e+02 4.801e+02, threshold=4.012e+02, percent-clipped=1.0 +2023-03-21 13:24:29,544 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 13:24:33,859 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 +2023-03-21 13:24:35,015 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 13:24:45,037 INFO [train.py:901] (1/2) Epoch 49, batch 250, loss[loss=0.1366, simple_loss=0.2225, pruned_loss=0.02533, over 6756.00 frames. ], tot_loss[loss=0.125, simple_loss=0.2062, pruned_loss=0.0219, over 1027003.28 frames. ], batch size: 107, lr: 3.38e-03, grad_scale: 4.0 +2023-03-21 13:24:48,101 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 13:24:51,497 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 +2023-03-21 13:24:54,207 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7318, 3.8988, 3.6127, 3.8560, 3.5059, 3.7401, 4.1166, 4.0913], + device='cuda:1'), covar=tensor([0.0233, 0.0160, 0.0251, 0.0179, 0.0297, 0.0388, 0.0205, 0.0184], + device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0131, 0.0126, 0.0129, 0.0116, 0.0104, 0.0102, 0.0105], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 13:24:56,045 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 +2023-03-21 13:24:57,254 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135827.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:25:08,048 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 13:25:11,182 INFO [train.py:901] (1/2) Epoch 49, batch 300, loss[loss=0.1545, simple_loss=0.2365, pruned_loss=0.03624, over 7313.00 frames. ], tot_loss[loss=0.1252, simple_loss=0.2064, pruned_loss=0.02202, over 1117657.45 frames. ], batch size: 49, lr: 3.38e-03, grad_scale: 4.0 +2023-03-21 13:25:11,554 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 +2023-03-21 13:25:17,397 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 13:25:18,891 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.312e+02 1.783e+02 2.054e+02 2.391e+02 4.462e+02, threshold=4.108e+02, percent-clipped=1.0 +2023-03-21 13:25:23,068 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=135875.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:25:37,116 INFO [train.py:901] (1/2) Epoch 49, batch 350, loss[loss=0.1509, simple_loss=0.2294, pruned_loss=0.0362, over 7314.00 frames. ], tot_loss[loss=0.126, simple_loss=0.2073, pruned_loss=0.02237, over 1190324.38 frames. ], batch size: 49, lr: 3.38e-03, grad_scale: 4.0 +2023-03-21 13:25:49,170 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=2.92 vs. limit=5.0 +2023-03-21 13:25:52,285 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 13:26:01,643 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-03-21 13:26:03,405 INFO [train.py:901] (1/2) Epoch 49, batch 400, loss[loss=0.1365, simple_loss=0.2177, pruned_loss=0.02767, over 7310.00 frames. ], tot_loss[loss=0.1258, simple_loss=0.2072, pruned_loss=0.02222, over 1248951.24 frames. ], batch size: 57, lr: 3.38e-03, grad_scale: 8.0 +2023-03-21 13:26:10,508 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.343e+02 1.801e+02 2.070e+02 2.366e+02 4.044e+02, threshold=4.140e+02, percent-clipped=0.0 +2023-03-21 13:26:18,113 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135982.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:26:20,102 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135986.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:26:24,162 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3772, 4.0988, 3.2944, 3.8815, 3.2997, 2.2209, 1.9317, 4.2946], + device='cuda:1'), covar=tensor([0.0044, 0.0074, 0.0166, 0.0079, 0.0161, 0.0623, 0.0732, 0.0046], + device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0094, 0.0117, 0.0100, 0.0134, 0.0135, 0.0132, 0.0107], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 13:26:24,184 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135994.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:26:32,946 INFO [train.py:901] (1/2) Epoch 49, batch 450, loss[loss=0.1303, simple_loss=0.206, pruned_loss=0.02732, over 7255.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.2073, pruned_loss=0.02227, over 1294107.24 frames. ], batch size: 47, lr: 3.38e-03, grad_scale: 8.0 +2023-03-21 13:26:40,079 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 13:26:40,541 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 13:26:54,490 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136043.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:26:59,373 INFO [train.py:901] (1/2) Epoch 49, batch 500, loss[loss=0.1303, simple_loss=0.2107, pruned_loss=0.02495, over 7285.00 frames. ], tot_loss[loss=0.126, simple_loss=0.2075, pruned_loss=0.02226, over 1327305.45 frames. ], batch size: 86, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:27:00,492 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136055.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:27:06,355 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.271e+02 1.831e+02 2.073e+02 2.387e+02 4.515e+02, threshold=4.145e+02, percent-clipped=2.0 +2023-03-21 13:27:11,628 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4171, 3.0665, 3.1025, 3.2862, 2.8989, 2.8476, 3.4576, 2.2700], + device='cuda:1'), covar=tensor([0.0717, 0.0795, 0.1104, 0.1024, 0.0879, 0.1353, 0.0969, 0.3365], + device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0338, 0.0268, 0.0347, 0.0277, 0.0284, 0.0342, 0.0233], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 13:27:12,943 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 13:27:13,966 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 13:27:14,942 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 13:27:16,950 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 13:27:21,498 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 13:27:24,757 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7887, 2.3780, 2.9567, 2.9394, 2.9096, 2.7659, 2.5260, 2.9819], + device='cuda:1'), covar=tensor([0.1408, 0.1035, 0.0915, 0.0973, 0.0808, 0.0968, 0.1673, 0.0923], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0072, 0.0054, 0.0053, 0.0053, 0.0052, 0.0071, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 13:27:25,081 INFO [train.py:901] (1/2) Epoch 49, batch 550, loss[loss=0.1308, simple_loss=0.2168, pruned_loss=0.02236, over 7281.00 frames. ], tot_loss[loss=0.1258, simple_loss=0.2073, pruned_loss=0.02212, over 1352909.79 frames. ], batch size: 77, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:27:27,922 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 +2023-03-21 13:27:28,619 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6040, 2.9467, 3.5802, 3.5965, 3.6700, 3.6805, 3.4511, 3.5641], + device='cuda:1'), covar=tensor([0.0029, 0.0124, 0.0034, 0.0031, 0.0029, 0.0029, 0.0071, 0.0050], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0075, 0.0062, 0.0060, 0.0057, 0.0064, 0.0049, 0.0083], + device='cuda:1'), out_proj_covar=tensor([8.6112e-05, 1.4729e-04, 1.0818e-04, 9.9591e-05, 9.3014e-05, 1.0717e-04, + 9.0443e-05, 1.4840e-04], device='cuda:1') +2023-03-21 13:27:32,972 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 13:27:38,239 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5532, 4.3415, 3.5864, 4.0336, 3.5796, 2.4235, 2.0907, 4.4563], + device='cuda:1'), covar=tensor([0.0046, 0.0061, 0.0133, 0.0071, 0.0138, 0.0557, 0.0681, 0.0048], + device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0094, 0.0117, 0.0100, 0.0135, 0.0135, 0.0132, 0.0107], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 13:27:41,230 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 13:27:44,246 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 13:27:45,396 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136142.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:27:50,831 INFO [train.py:901] (1/2) Epoch 49, batch 600, loss[loss=0.1012, simple_loss=0.1699, pruned_loss=0.01629, over 6476.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2077, pruned_loss=0.0225, over 1373152.43 frames. ], batch size: 28, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:27:52,402 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 13:27:53,501 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8107, 3.2637, 3.9068, 3.9340, 4.0552, 3.8596, 4.0207, 3.6714], + device='cuda:1'), covar=tensor([0.0059, 0.0184, 0.0048, 0.0050, 0.0047, 0.0058, 0.0061, 0.0080], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0075, 0.0062, 0.0060, 0.0057, 0.0064, 0.0050, 0.0083], + device='cuda:1'), out_proj_covar=tensor([8.6555e-05, 1.4797e-04, 1.0865e-04, 9.9767e-05, 9.3274e-05, 1.0764e-04, + 9.0685e-05, 1.4883e-04], device='cuda:1') +2023-03-21 13:27:57,841 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.300e+02 1.816e+02 2.119e+02 2.512e+02 4.639e+02, threshold=4.239e+02, percent-clipped=1.0 +2023-03-21 13:28:09,602 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 13:28:14,522 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 +2023-03-21 13:28:17,041 INFO [train.py:901] (1/2) Epoch 49, batch 650, loss[loss=0.1262, simple_loss=0.1998, pruned_loss=0.02631, over 7325.00 frames. ], tot_loss[loss=0.1263, simple_loss=0.2077, pruned_loss=0.02251, over 1390179.32 frames. ], batch size: 54, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:28:17,200 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136203.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:28:18,553 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 13:28:21,324 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5816, 2.1252, 2.6649, 2.6329, 2.6327, 2.5640, 2.2456, 2.7453], + device='cuda:1'), covar=tensor([0.1243, 0.1304, 0.1081, 0.1099, 0.0815, 0.0842, 0.1632, 0.0973], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0072, 0.0054, 0.0053, 0.0053, 0.0052, 0.0071, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 13:28:35,794 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 13:28:36,384 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0048, 4.3811, 4.6914, 4.6366, 4.6703, 4.5191, 4.9577, 4.4343], + device='cuda:1'), covar=tensor([0.0097, 0.0171, 0.0078, 0.0126, 0.0324, 0.0098, 0.0077, 0.0138], + device='cuda:1'), in_proj_covar=tensor([0.0103, 0.0110, 0.0111, 0.0096, 0.0187, 0.0116, 0.0112, 0.0122], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 13:28:41,222 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 +2023-03-21 13:28:42,960 INFO [train.py:901] (1/2) Epoch 49, batch 700, loss[loss=0.146, simple_loss=0.224, pruned_loss=0.03403, over 7249.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2077, pruned_loss=0.02253, over 1401085.53 frames. ], batch size: 55, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:28:44,597 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 13:28:50,198 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.282e+02 1.813e+02 2.063e+02 2.417e+02 3.262e+02, threshold=4.126e+02, percent-clipped=0.0 +2023-03-21 13:29:00,520 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136286.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:29:09,802 INFO [train.py:901] (1/2) Epoch 49, batch 750, loss[loss=0.1355, simple_loss=0.2172, pruned_loss=0.02688, over 7361.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.2071, pruned_loss=0.02215, over 1411708.06 frames. ], batch size: 65, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:29:10,318 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 13:29:10,858 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 13:29:19,939 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-03-21 13:29:20,303 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5680, 1.7899, 1.5140, 1.8007, 1.8746, 1.7287, 1.6779, 1.3355], + device='cuda:1'), covar=tensor([0.0286, 0.0161, 0.0418, 0.0195, 0.0154, 0.0224, 0.0181, 0.0216], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0042, 0.0040, 0.0041, 0.0039, 0.0039, 0.0040, 0.0050], + device='cuda:1'), out_proj_covar=tensor([4.7965e-05, 4.5987e-05, 4.4752e-05, 4.5333e-05, 4.3000e-05, 4.3112e-05, + 4.5205e-05, 5.4896e-05], device='cuda:1') +2023-03-21 13:29:23,856 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7273, 1.6531, 1.9832, 2.1784, 2.0508, 2.0440, 1.7074, 2.1811], + device='cuda:1'), covar=tensor([0.2941, 0.3564, 0.1344, 0.0781, 0.2708, 0.1283, 0.1850, 0.1746], + device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0087, 0.0079, 0.0069, 0.0071, 0.0070, 0.0111, 0.0072], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 13:29:24,733 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 13:29:25,787 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=136334.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:29:26,339 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136335.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:29:27,791 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136338.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:29:29,777 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 13:29:34,026 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136350.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:29:35,490 INFO [train.py:901] (1/2) Epoch 49, batch 800, loss[loss=0.1308, simple_loss=0.2166, pruned_loss=0.02254, over 7278.00 frames. ], tot_loss[loss=0.1252, simple_loss=0.2065, pruned_loss=0.02196, over 1416566.26 frames. ], batch size: 86, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:29:35,497 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 13:29:37,016 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 13:29:43,115 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 1.743e+02 2.098e+02 2.420e+02 5.303e+02, threshold=4.195e+02, percent-clipped=1.0 +2023-03-21 13:29:47,711 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 13:29:58,605 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136396.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:30:00,088 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0299, 4.2479, 3.9007, 4.2340, 3.9743, 4.1540, 4.5200, 4.4959], + device='cuda:1'), covar=tensor([0.0235, 0.0145, 0.0250, 0.0170, 0.0279, 0.0283, 0.0235, 0.0200], + device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0131, 0.0127, 0.0129, 0.0116, 0.0104, 0.0103, 0.0105], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 13:30:02,309 INFO [train.py:901] (1/2) Epoch 49, batch 850, loss[loss=0.1232, simple_loss=0.2083, pruned_loss=0.01901, over 7314.00 frames. ], tot_loss[loss=0.125, simple_loss=0.2064, pruned_loss=0.02179, over 1423445.48 frames. ], batch size: 59, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:30:07,273 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 13:30:07,277 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 13:30:12,381 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 13:30:15,414 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 13:30:28,108 INFO [train.py:901] (1/2) Epoch 49, batch 900, loss[loss=0.1245, simple_loss=0.2058, pruned_loss=0.02154, over 7306.00 frames. ], tot_loss[loss=0.1255, simple_loss=0.2071, pruned_loss=0.022, over 1425663.62 frames. ], batch size: 80, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:30:35,853 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.260e+02 1.732e+02 2.021e+02 2.349e+02 3.234e+02, threshold=4.042e+02, percent-clipped=0.0 +2023-03-21 13:30:51,561 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136498.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:30:53,933 INFO [train.py:901] (1/2) Epoch 49, batch 950, loss[loss=0.1269, simple_loss=0.2049, pruned_loss=0.02443, over 7227.00 frames. ], tot_loss[loss=0.1256, simple_loss=0.2071, pruned_loss=0.02202, over 1429960.75 frames. ], batch size: 45, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:30:55,025 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 13:31:18,888 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 13:31:19,893 INFO [train.py:901] (1/2) Epoch 49, batch 1000, loss[loss=0.1146, simple_loss=0.2002, pruned_loss=0.01451, over 7284.00 frames. ], tot_loss[loss=0.1261, simple_loss=0.2077, pruned_loss=0.02229, over 1433150.46 frames. ], batch size: 66, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:31:27,262 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3439, 2.2790, 2.4822, 2.1696, 2.5188, 2.3511, 2.1963, 1.8601], + device='cuda:1'), covar=tensor([0.0396, 0.0539, 0.0290, 0.0333, 0.0540, 0.0470, 0.0419, 0.0399], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0044, 0.0044, 0.0044, 0.0041, 0.0042, 0.0047, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 13:31:27,612 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.406e+02 1.764e+02 1.959e+02 2.305e+02 3.488e+02, threshold=3.918e+02, percent-clipped=0.0 +2023-03-21 13:31:40,794 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 13:31:46,171 INFO [train.py:901] (1/2) Epoch 49, batch 1050, loss[loss=0.137, simple_loss=0.2179, pruned_loss=0.02801, over 7235.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.2073, pruned_loss=0.02229, over 1434101.84 frames. ], batch size: 89, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:32:02,311 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 +2023-03-21 13:32:02,489 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 13:32:04,577 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136638.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:32:07,596 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 13:32:07,729 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136643.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:32:11,324 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136650.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:32:12,722 INFO [train.py:901] (1/2) Epoch 49, batch 1100, loss[loss=0.1444, simple_loss=0.2187, pruned_loss=0.03502, over 7328.00 frames. ], tot_loss[loss=0.1255, simple_loss=0.2069, pruned_loss=0.02204, over 1435416.07 frames. ], batch size: 75, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:32:19,842 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 8.990e+01 1.716e+02 2.031e+02 2.257e+02 3.350e+02, threshold=4.062e+02, percent-clipped=0.0 +2023-03-21 13:32:23,523 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0544, 3.4408, 2.9623, 3.5045, 3.3744, 2.9325, 3.2348, 3.2202], + device='cuda:1'), covar=tensor([0.1003, 0.0952, 0.1169, 0.0977, 0.0921, 0.0854, 0.0978, 0.1092], + device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0063, 0.0070, 0.0062, 0.0059, 0.0067, 0.0060, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:32:29,505 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=136686.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:32:31,934 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136691.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:32:32,817 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 +2023-03-21 13:32:35,507 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=136698.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:32:36,004 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 13:32:36,480 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 13:32:38,607 INFO [train.py:901] (1/2) Epoch 49, batch 1150, loss[loss=0.1212, simple_loss=0.2118, pruned_loss=0.01526, over 7278.00 frames. ], tot_loss[loss=0.126, simple_loss=0.2072, pruned_loss=0.02241, over 1435636.88 frames. ], batch size: 77, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:32:39,289 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136704.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:32:41,236 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.9771, 4.5476, 4.4190, 4.9782, 4.7919, 4.9220, 4.4464, 4.5572], + device='cuda:1'), covar=tensor([0.0900, 0.2577, 0.2374, 0.1212, 0.1042, 0.1130, 0.0813, 0.1125], + device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0404, 0.0303, 0.0319, 0.0240, 0.0379, 0.0239, 0.0287], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 13:32:48,849 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 13:32:49,866 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 13:32:58,803 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5507, 1.7659, 1.5387, 1.7321, 1.8067, 1.7213, 1.7007, 1.4109], + device='cuda:1'), covar=tensor([0.0212, 0.0159, 0.0333, 0.0182, 0.0128, 0.0173, 0.0167, 0.0208], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0042, 0.0040, 0.0041, 0.0039, 0.0039, 0.0040, 0.0050], + device='cuda:1'), out_proj_covar=tensor([4.8064e-05, 4.5921e-05, 4.4661e-05, 4.4997e-05, 4.3067e-05, 4.2808e-05, + 4.5085e-05, 5.4998e-05], device='cuda:1') +2023-03-21 13:33:04,801 INFO [train.py:901] (1/2) Epoch 49, batch 1200, loss[loss=0.1368, simple_loss=0.2245, pruned_loss=0.02456, over 7273.00 frames. ], tot_loss[loss=0.1258, simple_loss=0.2071, pruned_loss=0.0223, over 1437123.43 frames. ], batch size: 77, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:33:11,771 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 1.907e+02 2.162e+02 2.451e+02 4.461e+02, threshold=4.324e+02, percent-clipped=2.0 +2023-03-21 13:33:22,651 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 13:33:27,192 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0842, 3.4696, 4.1123, 4.1561, 4.2044, 4.1705, 4.1178, 4.0105], + device='cuda:1'), covar=tensor([0.0036, 0.0120, 0.0032, 0.0031, 0.0027, 0.0030, 0.0035, 0.0054], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0076, 0.0063, 0.0060, 0.0058, 0.0064, 0.0050, 0.0084], + device='cuda:1'), out_proj_covar=tensor([8.7661e-05, 1.4840e-04, 1.0851e-04, 1.0030e-04, 9.4357e-05, 1.0794e-04, + 9.1547e-05, 1.5036e-04], device='cuda:1') +2023-03-21 13:33:28,205 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136798.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:33:30,890 INFO [train.py:901] (1/2) Epoch 49, batch 1250, loss[loss=0.1195, simple_loss=0.2031, pruned_loss=0.01798, over 7216.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.207, pruned_loss=0.02239, over 1437613.54 frames. ], batch size: 50, lr: 3.37e-03, grad_scale: 8.0 +2023-03-21 13:33:32,527 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136806.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:33:36,589 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8075, 3.1883, 2.7347, 3.0517, 2.9128, 2.7041, 2.9211, 2.9457], + device='cuda:1'), covar=tensor([0.0625, 0.0422, 0.1137, 0.0799, 0.0957, 0.0682, 0.0761, 0.0887], + device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0062, 0.0070, 0.0062, 0.0058, 0.0066, 0.0059, 0.0056], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:33:47,276 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 13:33:51,274 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 13:33:52,754 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 13:33:53,307 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=136846.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:33:56,773 INFO [train.py:901] (1/2) Epoch 49, batch 1300, loss[loss=0.1334, simple_loss=0.2127, pruned_loss=0.02702, over 7247.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.2068, pruned_loss=0.02234, over 1438259.54 frames. ], batch size: 55, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:33:58,993 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7559, 2.8712, 2.1227, 3.1900, 2.4036, 2.7563, 1.3844, 2.1384], + device='cuda:1'), covar=tensor([0.0719, 0.0987, 0.2999, 0.0937, 0.0657, 0.0732, 0.4479, 0.2039], + device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0258, 0.0275, 0.0267, 0.0264, 0.0262, 0.0226, 0.0254], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:34:03,864 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.273e+02 1.719e+02 1.984e+02 2.382e+02 4.139e+02, threshold=3.968e+02, percent-clipped=0.0 +2023-03-21 13:34:04,056 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136867.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:34:16,106 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 13:34:18,609 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 13:34:22,166 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 13:34:22,645 INFO [train.py:901] (1/2) Epoch 49, batch 1350, loss[loss=0.1229, simple_loss=0.2123, pruned_loss=0.01677, over 7290.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.2069, pruned_loss=0.02227, over 1439302.53 frames. ], batch size: 66, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:34:33,356 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 13:34:38,315 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-03-21 13:34:48,549 INFO [train.py:901] (1/2) Epoch 49, batch 1400, loss[loss=0.1275, simple_loss=0.2119, pruned_loss=0.02158, over 7328.00 frames. ], tot_loss[loss=0.126, simple_loss=0.2074, pruned_loss=0.02232, over 1438775.02 frames. ], batch size: 49, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:34:56,149 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.332e+02 1.767e+02 1.984e+02 2.313e+02 6.157e+02, threshold=3.967e+02, percent-clipped=1.0 +2023-03-21 13:35:05,316 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 13:35:08,440 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136991.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:35:13,073 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136999.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:35:15,393 INFO [train.py:901] (1/2) Epoch 49, batch 1450, loss[loss=0.121, simple_loss=0.2084, pruned_loss=0.01676, over 7284.00 frames. ], tot_loss[loss=0.1263, simple_loss=0.2079, pruned_loss=0.02231, over 1440629.18 frames. ], batch size: 77, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:35:18,539 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3989, 2.5001, 2.8069, 2.4218, 2.7174, 2.5739, 2.3791, 2.1358], + device='cuda:1'), covar=tensor([0.0489, 0.0672, 0.0235, 0.0292, 0.0425, 0.0445, 0.0350, 0.0446], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0043, 0.0043, 0.0043, 0.0041, 0.0041, 0.0047, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 13:35:29,945 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 13:35:33,541 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=137039.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:35:40,597 INFO [train.py:901] (1/2) Epoch 49, batch 1500, loss[loss=0.1182, simple_loss=0.2001, pruned_loss=0.01814, over 7283.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.2075, pruned_loss=0.02191, over 1441445.47 frames. ], batch size: 57, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:35:46,934 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 13:35:48,437 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.668e+02 1.982e+02 2.289e+02 3.254e+02, threshold=3.964e+02, percent-clipped=0.0 +2023-03-21 13:36:07,420 INFO [train.py:901] (1/2) Epoch 49, batch 1550, loss[loss=0.1302, simple_loss=0.2082, pruned_loss=0.02611, over 7285.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.2075, pruned_loss=0.02213, over 1443779.35 frames. ], batch size: 57, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:36:12,014 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 13:36:12,351 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.83 vs. limit=2.0 +2023-03-21 13:36:33,347 INFO [train.py:901] (1/2) Epoch 49, batch 1600, loss[loss=0.1232, simple_loss=0.211, pruned_loss=0.01771, over 7265.00 frames. ], tot_loss[loss=0.1261, simple_loss=0.2076, pruned_loss=0.02233, over 1444271.26 frames. ], batch size: 64, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:36:37,935 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=137162.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:36:40,361 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.807e+02 2.128e+02 2.612e+02 4.654e+02, threshold=4.257e+02, percent-clipped=2.0 +2023-03-21 13:36:44,099 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 13:36:45,109 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 13:36:48,159 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 13:36:56,504 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-03-21 13:36:57,732 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 13:36:59,563 INFO [train.py:901] (1/2) Epoch 49, batch 1650, loss[loss=0.1251, simple_loss=0.2091, pruned_loss=0.02059, over 7312.00 frames. ], tot_loss[loss=0.126, simple_loss=0.2072, pruned_loss=0.02238, over 1443370.98 frames. ], batch size: 59, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:37:02,088 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 13:37:04,247 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1504, 3.1457, 3.3633, 3.3227, 3.2398, 3.1511, 2.8537, 3.3027], + device='cuda:1'), covar=tensor([0.1646, 0.0586, 0.0958, 0.0990, 0.0859, 0.0848, 0.1725, 0.1523], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0073, 0.0054, 0.0053, 0.0054, 0.0052, 0.0071, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 13:37:10,655 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 13:37:25,329 INFO [train.py:901] (1/2) Epoch 49, batch 1700, loss[loss=0.1355, simple_loss=0.2109, pruned_loss=0.03009, over 7274.00 frames. ], tot_loss[loss=0.1255, simple_loss=0.2065, pruned_loss=0.02224, over 1441946.69 frames. ], batch size: 66, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:37:27,588 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.6478, 4.0503, 2.8125, 4.2021, 3.4583, 4.0040, 1.9229, 2.6800], + device='cuda:1'), covar=tensor([0.0484, 0.0977, 0.2984, 0.0476, 0.0451, 0.0718, 0.4508, 0.2381], + device='cuda:1'), in_proj_covar=tensor([0.0274, 0.0257, 0.0273, 0.0266, 0.0264, 0.0262, 0.0225, 0.0252], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:37:28,500 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 13:37:33,086 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.251e+02 1.788e+02 2.140e+02 2.383e+02 3.723e+02, threshold=4.279e+02, percent-clipped=0.0 +2023-03-21 13:37:33,640 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 13:37:43,288 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 13:37:49,456 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137299.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:37:51,370 INFO [train.py:901] (1/2) Epoch 49, batch 1750, loss[loss=0.1056, simple_loss=0.1887, pruned_loss=0.01121, over 7127.00 frames. ], tot_loss[loss=0.1256, simple_loss=0.2066, pruned_loss=0.02228, over 1439882.40 frames. ], batch size: 41, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:38:02,550 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3166, 4.7809, 4.8867, 4.7722, 4.7233, 4.3156, 4.8747, 4.6715], + device='cuda:1'), covar=tensor([0.0441, 0.0363, 0.0325, 0.0480, 0.0349, 0.0448, 0.0314, 0.0470], + device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0273, 0.0214, 0.0213, 0.0164, 0.0241, 0.0223, 0.0153], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 13:38:07,005 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 13:38:08,587 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 13:38:13,857 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3074, 4.7327, 4.8170, 4.7252, 4.7208, 4.2959, 4.8099, 4.6571], + device='cuda:1'), covar=tensor([0.0455, 0.0344, 0.0334, 0.0488, 0.0344, 0.0401, 0.0331, 0.0464], + device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0273, 0.0214, 0.0213, 0.0164, 0.0242, 0.0223, 0.0153], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 13:38:14,843 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=137347.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:38:17,824 INFO [train.py:901] (1/2) Epoch 49, batch 1800, loss[loss=0.1253, simple_loss=0.2103, pruned_loss=0.02016, over 7308.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.207, pruned_loss=0.02217, over 1441569.75 frames. ], batch size: 80, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:38:25,031 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 1.722e+02 1.981e+02 2.424e+02 4.564e+02, threshold=3.962e+02, percent-clipped=1.0 +2023-03-21 13:38:29,543 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 13:38:42,619 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 13:38:44,216 INFO [train.py:901] (1/2) Epoch 49, batch 1850, loss[loss=0.1035, simple_loss=0.1737, pruned_loss=0.0167, over 5889.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.207, pruned_loss=0.02218, over 1440312.79 frames. ], batch size: 25, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:38:52,779 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 13:39:09,627 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 13:39:10,090 INFO [train.py:901] (1/2) Epoch 49, batch 1900, loss[loss=0.1327, simple_loss=0.2175, pruned_loss=0.02393, over 7247.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.2073, pruned_loss=0.02226, over 1440835.34 frames. ], batch size: 89, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:39:14,754 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137462.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:39:14,977 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 +2023-03-21 13:39:17,245 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.231e+02 1.672e+02 1.883e+02 2.145e+02 4.351e+02, threshold=3.765e+02, percent-clipped=1.0 +2023-03-21 13:39:35,151 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 13:39:36,114 INFO [train.py:901] (1/2) Epoch 49, batch 1950, loss[loss=0.1313, simple_loss=0.211, pruned_loss=0.0258, over 7290.00 frames. ], tot_loss[loss=0.1256, simple_loss=0.2068, pruned_loss=0.02223, over 1439570.98 frames. ], batch size: 66, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:39:39,421 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2743, 2.8424, 2.1484, 3.0830, 3.0049, 3.1611, 2.8436, 2.6679], + device='cuda:1'), covar=tensor([0.2427, 0.1212, 0.4050, 0.0515, 0.0303, 0.0280, 0.0448, 0.0514], + device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0231, 0.0242, 0.0252, 0.0203, 0.0206, 0.0219, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:39:39,840 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=137510.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:39:46,589 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 13:39:51,242 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 13:39:51,765 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 13:39:52,886 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0602, 4.5801, 4.4023, 5.0009, 4.8680, 4.9224, 4.4410, 4.5769], + device='cuda:1'), covar=tensor([0.0766, 0.2319, 0.2137, 0.1118, 0.0781, 0.1137, 0.0830, 0.1108], + device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0408, 0.0304, 0.0317, 0.0239, 0.0381, 0.0240, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 13:40:02,186 INFO [train.py:901] (1/2) Epoch 49, batch 2000, loss[loss=0.1244, simple_loss=0.2069, pruned_loss=0.02093, over 7356.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.207, pruned_loss=0.02221, over 1439599.35 frames. ], batch size: 61, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:40:07,110 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 13:40:07,740 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2178, 3.9296, 3.7799, 3.8987, 3.9366, 3.7863, 4.1452, 3.6380], + device='cuda:1'), covar=tensor([0.0151, 0.0173, 0.0161, 0.0184, 0.0431, 0.0143, 0.0143, 0.0196], + device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0107, 0.0110, 0.0095, 0.0184, 0.0114, 0.0110, 0.0120], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 13:40:09,125 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.480e+02 1.776e+02 2.003e+02 2.327e+02 4.090e+02, threshold=4.007e+02, percent-clipped=2.0 +2023-03-21 13:40:18,422 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 13:40:27,037 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 13:40:28,912 INFO [train.py:901] (1/2) Epoch 49, batch 2050, loss[loss=0.1197, simple_loss=0.207, pruned_loss=0.01618, over 7302.00 frames. ], tot_loss[loss=0.1258, simple_loss=0.207, pruned_loss=0.02224, over 1439091.12 frames. ], batch size: 83, lr: 3.36e-03, grad_scale: 8.0 +2023-03-21 13:40:51,635 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 +2023-03-21 13:40:54,343 INFO [train.py:901] (1/2) Epoch 49, batch 2100, loss[loss=0.1257, simple_loss=0.205, pruned_loss=0.0232, over 7354.00 frames. ], tot_loss[loss=0.1255, simple_loss=0.2072, pruned_loss=0.0219, over 1439608.37 frames. ], batch size: 63, lr: 3.35e-03, grad_scale: 8.0 +2023-03-21 13:41:00,108 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 13:41:02,071 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.230e+02 1.716e+02 1.929e+02 2.374e+02 1.114e+03, threshold=3.858e+02, percent-clipped=2.0 +2023-03-21 13:41:03,086 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 13:41:20,853 INFO [train.py:901] (1/2) Epoch 49, batch 2150, loss[loss=0.1276, simple_loss=0.2051, pruned_loss=0.02505, over 7265.00 frames. ], tot_loss[loss=0.1261, simple_loss=0.2077, pruned_loss=0.02231, over 1441043.86 frames. ], batch size: 55, lr: 3.35e-03, grad_scale: 8.0 +2023-03-21 13:41:43,635 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=2.19 vs. limit=2.0 +2023-03-21 13:41:46,881 INFO [train.py:901] (1/2) Epoch 49, batch 2200, loss[loss=0.1483, simple_loss=0.2281, pruned_loss=0.03422, over 7268.00 frames. ], tot_loss[loss=0.1261, simple_loss=0.2078, pruned_loss=0.0222, over 1442701.73 frames. ], batch size: 64, lr: 3.35e-03, grad_scale: 8.0 +2023-03-21 13:41:49,967 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 13:41:54,584 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 1.769e+02 2.125e+02 2.372e+02 3.697e+02, threshold=4.249e+02, percent-clipped=0.0 +2023-03-21 13:41:55,384 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.18 vs. limit=5.0 +2023-03-21 13:42:06,031 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6085, 1.7896, 1.5788, 1.7217, 1.7855, 1.7985, 1.7523, 1.5145], + device='cuda:1'), covar=tensor([0.0250, 0.0194, 0.0297, 0.0173, 0.0167, 0.0166, 0.0180, 0.0248], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0042, 0.0040, 0.0041, 0.0039, 0.0038, 0.0040, 0.0050], + device='cuda:1'), out_proj_covar=tensor([4.7483e-05, 4.5993e-05, 4.5143e-05, 4.4883e-05, 4.3106e-05, 4.2571e-05, + 4.5081e-05, 5.4860e-05], device='cuda:1') +2023-03-21 13:42:13,234 INFO [train.py:901] (1/2) Epoch 49, batch 2250, loss[loss=0.1177, simple_loss=0.2026, pruned_loss=0.0164, over 7302.00 frames. ], tot_loss[loss=0.1261, simple_loss=0.2077, pruned_loss=0.02226, over 1442864.84 frames. ], batch size: 80, lr: 3.35e-03, grad_scale: 16.0 +2023-03-21 13:42:13,358 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4002, 4.0303, 3.9760, 4.0862, 4.0562, 3.9632, 4.2927, 3.7553], + device='cuda:1'), covar=tensor([0.0132, 0.0160, 0.0125, 0.0154, 0.0411, 0.0131, 0.0147, 0.0182], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0106, 0.0109, 0.0093, 0.0182, 0.0113, 0.0109, 0.0119], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 13:42:20,516 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3429, 3.1131, 3.1040, 3.2672, 2.8355, 2.8865, 3.4302, 2.2400], + device='cuda:1'), covar=tensor([0.0758, 0.0796, 0.0991, 0.0921, 0.0891, 0.1257, 0.1135, 0.3313], + device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0338, 0.0270, 0.0349, 0.0280, 0.0284, 0.0344, 0.0235], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 13:42:24,919 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 13:42:25,426 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 13:42:31,839 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1038, 2.4254, 1.8818, 2.7532, 2.8451, 2.7501, 2.5453, 2.6234], + device='cuda:1'), covar=tensor([0.2409, 0.1256, 0.4008, 0.0998, 0.0588, 0.0553, 0.0604, 0.0640], + device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0229, 0.0240, 0.0252, 0.0202, 0.0204, 0.0217, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:42:38,376 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6372, 1.6499, 1.6616, 1.9389, 1.6273, 1.9186, 1.6064, 1.9427], + device='cuda:1'), covar=tensor([0.2399, 0.3133, 0.2075, 0.0969, 0.1408, 0.1183, 0.2799, 0.1230], + device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0087, 0.0079, 0.0071, 0.0070, 0.0070, 0.0112, 0.0072], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 13:42:38,728 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 13:42:39,225 INFO [train.py:901] (1/2) Epoch 49, batch 2300, loss[loss=0.1184, simple_loss=0.2067, pruned_loss=0.01504, over 7329.00 frames. ], tot_loss[loss=0.1262, simple_loss=0.2081, pruned_loss=0.02216, over 1439562.47 frames. ], batch size: 44, lr: 3.35e-03, grad_scale: 8.0 +2023-03-21 13:42:47,627 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.190e+02 1.772e+02 2.010e+02 2.358e+02 5.104e+02, threshold=4.021e+02, percent-clipped=1.0 +2023-03-21 13:42:52,762 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-03-21 13:43:00,119 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137892.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:43:01,577 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137895.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:43:04,554 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0228, 3.3107, 4.0208, 4.0732, 4.1431, 4.1011, 4.2481, 3.9850], + device='cuda:1'), covar=tensor([0.0034, 0.0130, 0.0034, 0.0033, 0.0027, 0.0029, 0.0033, 0.0045], + device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0077, 0.0063, 0.0061, 0.0058, 0.0064, 0.0051, 0.0084], + device='cuda:1'), out_proj_covar=tensor([8.8834e-05, 1.5046e-04, 1.0937e-04, 1.0129e-04, 9.4631e-05, 1.0771e-04, + 9.2240e-05, 1.5075e-04], device='cuda:1') +2023-03-21 13:43:05,462 INFO [train.py:901] (1/2) Epoch 49, batch 2350, loss[loss=0.129, simple_loss=0.2115, pruned_loss=0.02325, over 7323.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.2076, pruned_loss=0.02203, over 1439768.65 frames. ], batch size: 75, lr: 3.35e-03, grad_scale: 8.0 +2023-03-21 13:43:10,749 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-03-21 13:43:27,517 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 13:43:28,498 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-03-21 13:43:32,245 INFO [train.py:901] (1/2) Epoch 49, batch 2400, loss[loss=0.1328, simple_loss=0.2147, pruned_loss=0.02542, over 7283.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.2074, pruned_loss=0.02222, over 1439387.16 frames. ], batch size: 77, lr: 3.35e-03, grad_scale: 8.0 +2023-03-21 13:43:32,396 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137953.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:43:33,823 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137956.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:43:34,168 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 13:43:39,737 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.383e+02 1.856e+02 2.125e+02 2.479e+02 3.936e+02, threshold=4.250e+02, percent-clipped=0.0 +2023-03-21 13:43:44,337 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 13:43:47,471 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 13:43:58,057 INFO [train.py:901] (1/2) Epoch 49, batch 2450, loss[loss=0.1626, simple_loss=0.2464, pruned_loss=0.03945, over 6651.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.2074, pruned_loss=0.02219, over 1439089.84 frames. ], batch size: 106, lr: 3.35e-03, grad_scale: 8.0 +2023-03-21 13:44:07,360 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.4903, 3.6405, 3.5004, 3.6173, 3.4317, 3.5257, 3.8960, 3.8975], + device='cuda:1'), covar=tensor([0.0229, 0.0158, 0.0223, 0.0169, 0.0296, 0.0328, 0.0204, 0.0170], + device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0130, 0.0124, 0.0127, 0.0115, 0.0103, 0.0101, 0.0103], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 13:44:14,544 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 13:44:24,608 INFO [train.py:901] (1/2) Epoch 49, batch 2500, loss[loss=0.1209, simple_loss=0.1986, pruned_loss=0.0216, over 7366.00 frames. ], tot_loss[loss=0.126, simple_loss=0.2075, pruned_loss=0.02228, over 1439445.86 frames. ], batch size: 51, lr: 3.35e-03, grad_scale: 8.0 +2023-03-21 13:44:32,197 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.701e+02 1.944e+02 2.363e+02 6.757e+02, threshold=3.888e+02, percent-clipped=1.0 +2023-03-21 13:44:39,635 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 13:44:43,501 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-03-21 13:44:50,360 INFO [train.py:901] (1/2) Epoch 49, batch 2550, loss[loss=0.1267, simple_loss=0.2104, pruned_loss=0.02148, over 7313.00 frames. ], tot_loss[loss=0.1261, simple_loss=0.2073, pruned_loss=0.02241, over 1437632.34 frames. ], batch size: 49, lr: 3.35e-03, grad_scale: 8.0 +2023-03-21 13:44:52,048 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2480, 2.6415, 2.0121, 2.8385, 2.8340, 2.8084, 2.5552, 2.6713], + device='cuda:1'), covar=tensor([0.2193, 0.1282, 0.3945, 0.0595, 0.0356, 0.0305, 0.0434, 0.0425], + device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0230, 0.0241, 0.0253, 0.0203, 0.0206, 0.0219, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:44:56,053 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0778, 2.7977, 1.8407, 3.3781, 3.4657, 3.3400, 3.1757, 3.1414], + device='cuda:1'), covar=tensor([0.2971, 0.1102, 0.5104, 0.0534, 0.0400, 0.0293, 0.0562, 0.0460], + device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0230, 0.0242, 0.0253, 0.0204, 0.0206, 0.0219, 0.0227], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:45:03,282 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9042, 3.2107, 2.2279, 3.2920, 2.4695, 2.9951, 1.4259, 2.3945], + device='cuda:1'), covar=tensor([0.0599, 0.1030, 0.3188, 0.0801, 0.0577, 0.0786, 0.4956, 0.1991], + device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0256, 0.0273, 0.0265, 0.0264, 0.0261, 0.0225, 0.0250], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:45:16,197 INFO [train.py:901] (1/2) Epoch 49, batch 2600, loss[loss=0.1371, simple_loss=0.2164, pruned_loss=0.02889, over 7287.00 frames. ], tot_loss[loss=0.126, simple_loss=0.2073, pruned_loss=0.02233, over 1438921.66 frames. ], batch size: 47, lr: 3.35e-03, grad_scale: 8.0 +2023-03-21 13:45:23,532 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 1.730e+02 2.027e+02 2.296e+02 4.292e+02, threshold=4.053e+02, percent-clipped=1.0 +2023-03-21 13:45:28,189 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.7992, 3.0817, 3.8591, 3.8369, 3.8772, 3.8697, 3.9436, 3.7871], + device='cuda:1'), covar=tensor([0.0040, 0.0141, 0.0031, 0.0037, 0.0031, 0.0033, 0.0046, 0.0049], + device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0076, 0.0063, 0.0061, 0.0058, 0.0064, 0.0050, 0.0083], + device='cuda:1'), out_proj_covar=tensor([8.8418e-05, 1.4930e-04, 1.0870e-04, 1.0079e-04, 9.4547e-05, 1.0698e-04, + 9.1802e-05, 1.4919e-04], device='cuda:1') +2023-03-21 13:45:28,341 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-03-21 13:45:41,142 INFO [train.py:901] (1/2) Epoch 49, batch 2650, loss[loss=0.1504, simple_loss=0.2261, pruned_loss=0.03732, over 7128.00 frames. ], tot_loss[loss=0.126, simple_loss=0.2077, pruned_loss=0.02212, over 1441828.54 frames. ], batch size: 98, lr: 3.35e-03, grad_scale: 8.0 +2023-03-21 13:46:02,200 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.3754, 3.0187, 3.1538, 3.2160, 2.9432, 3.0084, 3.3942, 2.3294], + device='cuda:1'), covar=tensor([0.0577, 0.0578, 0.0982, 0.0793, 0.0784, 0.1153, 0.0818, 0.3138], + device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0334, 0.0268, 0.0347, 0.0277, 0.0282, 0.0342, 0.0233], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 13:46:04,116 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138248.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:46:05,608 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138251.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:46:06,534 INFO [train.py:901] (1/2) Epoch 49, batch 2700, loss[loss=0.123, simple_loss=0.2123, pruned_loss=0.0168, over 7332.00 frames. ], tot_loss[loss=0.1258, simple_loss=0.208, pruned_loss=0.02186, over 1443461.64 frames. ], batch size: 75, lr: 3.35e-03, grad_scale: 8.0 +2023-03-21 13:46:13,884 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.373e+02 1.750e+02 2.007e+02 2.310e+02 3.758e+02, threshold=4.014e+02, percent-clipped=0.0 +2023-03-21 13:46:14,999 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138270.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:46:31,353 INFO [train.py:901] (1/2) Epoch 49, batch 2750, loss[loss=0.1361, simple_loss=0.2174, pruned_loss=0.02743, over 7287.00 frames. ], tot_loss[loss=0.1256, simple_loss=0.2076, pruned_loss=0.02186, over 1442845.31 frames. ], batch size: 66, lr: 3.35e-03, grad_scale: 8.0 +2023-03-21 13:46:40,790 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138322.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:46:45,769 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138331.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:46:56,604 INFO [train.py:901] (1/2) Epoch 49, batch 2800, loss[loss=0.1248, simple_loss=0.2086, pruned_loss=0.02048, over 7272.00 frames. ], tot_loss[loss=0.126, simple_loss=0.2077, pruned_loss=0.02218, over 1442155.83 frames. ], batch size: 70, lr: 3.35e-03, grad_scale: 8.0 +2023-03-21 13:47:04,003 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.824e+02 2.150e+02 2.654e+02 5.945e+02, threshold=4.299e+02, percent-clipped=1.0 +2023-03-21 13:47:19,954 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0445-102620-0_sp0.9 from training. Duration: 13.3300625 +2023-03-21 13:47:20,006 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0343W0353-107668-0_sp0.9 from training. Duration: 12.0068125 +2023-03-21 13:47:20,031 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0_sp0.9 from training. Duration: 13.7855625 +2023-03-21 13:47:20,050 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0322-35834-0_sp0.9 from training. Duration: 12.7411875 +2023-03-21 13:47:20,218 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp1.1 from training. Duration: 13.21025 +2023-03-21 13:47:20,234 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0174W0255-47639-0_sp0.9 from training. Duration: 12.394375 +2023-03-21 13:47:20,244 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0431-52838-0_sp0.9 from training. Duration: 12.390125 +2023-03-21 13:47:20,341 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0123-40756-0_sp0.9 from training. Duration: 12.1100625 +2023-03-21 13:47:20,746 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0233-86222-0_sp0.9 from training. Duration: 13.103375 +2023-03-21 13:47:20,762 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0 from training. Duration: 12.021 +2023-03-21 13:47:20,778 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0495-41823-0_sp0.9 from training. Duration: 12.204875 +2023-03-21 13:47:20,936 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0107W0142-64458-0_sp0.9 from training. Duration: 12.61225 +2023-03-21 13:47:20,942 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0 from training. Duration: 12.1599375 +2023-03-21 13:47:20,949 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0466-31670-0_sp0.9 from training. Duration: 12.250125 +2023-03-21 13:47:21,083 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0126W0130-17071-0_sp0.9 from training. Duration: 12.6745 +2023-03-21 13:47:21,086 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0355-15040-0_sp0.9 from training. Duration: 13.2076875 +2023-03-21 13:47:21,143 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0 from training. Duration: 12.3199375 +2023-03-21 13:47:25,317 INFO [train.py:901] (1/2) Epoch 50, batch 0, loss[loss=0.123, simple_loss=0.2084, pruned_loss=0.01883, over 7292.00 frames. ], tot_loss[loss=0.123, simple_loss=0.2084, pruned_loss=0.01883, over 7292.00 frames. ], batch size: 68, lr: 3.31e-03, grad_scale: 8.0 +2023-03-21 13:47:25,317 INFO [train.py:926] (1/2) Computing validation loss +2023-03-21 13:47:37,244 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3231, 4.0571, 3.7428, 4.4240, 4.0974, 4.3954, 4.2066, 4.0879], + device='cuda:1'), covar=tensor([0.0856, 0.2310, 0.2139, 0.1160, 0.0870, 0.1055, 0.0575, 0.1051], + device='cuda:1'), in_proj_covar=tensor([0.0275, 0.0409, 0.0304, 0.0321, 0.0243, 0.0382, 0.0240, 0.0291], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 13:47:51,867 INFO [train.py:935] (1/2) Epoch 50, validation: loss=0.1661, simple_loss=0.2589, pruned_loss=0.03663, over 1622729.00 frames. +2023-03-21 13:47:51,867 INFO [train.py:936] (1/2) Maximum memory allocated so far is 12682MB +2023-03-21 13:47:55,071 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138383.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:47:58,491 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0472-52669-0_sp0.9 from training. Duration: 12.9869375 +2023-03-21 13:48:09,994 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0377-52839-0_sp0.9 from training. Duration: 12.7031875 +2023-03-21 13:48:17,097 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0214-86176-0_sp0.9 from training. Duration: 13.6888125 +2023-03-21 13:48:17,612 INFO [train.py:901] (1/2) Epoch 50, batch 50, loss[loss=0.1216, simple_loss=0.21, pruned_loss=0.01664, over 7381.00 frames. ], tot_loss[loss=0.1282, simple_loss=0.21, pruned_loss=0.02318, over 325927.74 frames. ], batch size: 51, lr: 3.31e-03, grad_scale: 8.0 +2023-03-21 13:48:19,177 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0712W0259-53832-0_sp0.9 from training. Duration: 12.17325 +2023-03-21 13:48:22,135 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0468-86267-0_sp0.9 from training. Duration: 13.0943125 +2023-03-21 13:48:25,645 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 +2023-03-21 13:48:39,555 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.230e+02 1.699e+02 2.026e+02 2.461e+02 4.880e+02, threshold=4.052e+02, percent-clipped=2.0 +2023-03-21 13:48:40,592 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0127-86359-0_sp0.9 from training. Duration: 12.76 +2023-03-21 13:48:41,117 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0363-35919-0 from training. Duration: 12.407 +2023-03-21 13:48:44,159 INFO [train.py:901] (1/2) Epoch 50, batch 100, loss[loss=0.135, simple_loss=0.2213, pruned_loss=0.02431, over 7122.00 frames. ], tot_loss[loss=0.1262, simple_loss=0.2086, pruned_loss=0.02192, over 572018.37 frames. ], batch size: 98, lr: 3.31e-03, grad_scale: 8.0 +2023-03-21 13:49:10,093 INFO [train.py:901] (1/2) Epoch 50, batch 150, loss[loss=0.1387, simple_loss=0.2141, pruned_loss=0.03165, over 7216.00 frames. ], tot_loss[loss=0.1249, simple_loss=0.2072, pruned_loss=0.02129, over 765893.93 frames. ], batch size: 45, lr: 3.31e-03, grad_scale: 8.0 +2023-03-21 13:49:20,604 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138548.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:49:22,681 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138551.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:49:31,278 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.281e+02 1.689e+02 1.944e+02 2.232e+02 3.987e+02, threshold=3.888e+02, percent-clipped=0.0 +2023-03-21 13:49:35,862 INFO [train.py:901] (1/2) Epoch 50, batch 200, loss[loss=0.1668, simple_loss=0.2432, pruned_loss=0.04522, over 6779.00 frames. ], tot_loss[loss=0.1265, simple_loss=0.2086, pruned_loss=0.02218, over 916213.67 frames. ], batch size: 106, lr: 3.31e-03, grad_scale: 8.0 +2023-03-21 13:49:39,457 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0_sp0.9 from training. Duration: 13.503375 +2023-03-21 13:49:40,954 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.1010, 4.6177, 4.3306, 5.0717, 4.7525, 4.9995, 4.4677, 4.6821], + device='cuda:1'), covar=tensor([0.0800, 0.2412, 0.2452, 0.1045, 0.1085, 0.1173, 0.0785, 0.1212], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0409, 0.0305, 0.0321, 0.0244, 0.0384, 0.0241, 0.0292], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 13:49:43,934 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0480-86356-0_sp0.9 from training. Duration: 12.22225 +2023-03-21 13:49:45,502 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=138596.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:49:46,954 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=138599.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:49:50,859 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0 from training. Duration: 12.3249375 +2023-03-21 13:50:01,374 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138626.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:50:01,801 INFO [train.py:901] (1/2) Epoch 50, batch 250, loss[loss=0.1161, simple_loss=0.1899, pruned_loss=0.02109, over 7203.00 frames. ], tot_loss[loss=0.126, simple_loss=0.208, pruned_loss=0.02197, over 1033357.79 frames. ], batch size: 39, lr: 3.31e-03, grad_scale: 8.0 +2023-03-21 13:50:03,339 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0 from training. Duration: 12.185 +2023-03-21 13:50:22,912 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.763e+02 2.018e+02 2.436e+02 6.098e+02, threshold=4.036e+02, percent-clipped=4.0 +2023-03-21 13:50:23,495 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.3646, 4.8763, 4.6826, 5.3140, 5.1099, 5.2409, 4.7381, 4.9650], + device='cuda:1'), covar=tensor([0.0764, 0.2373, 0.2112, 0.1005, 0.0852, 0.1051, 0.0724, 0.0999], + device='cuda:1'), in_proj_covar=tensor([0.0276, 0.0409, 0.0303, 0.0321, 0.0242, 0.0383, 0.0241, 0.0290], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 13:50:24,930 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0284-35956-0_sp0.9 from training. Duration: 12.4655625 +2023-03-21 13:50:27,515 INFO [train.py:901] (1/2) Epoch 50, batch 300, loss[loss=0.121, simple_loss=0.2072, pruned_loss=0.01736, over 7345.00 frames. ], tot_loss[loss=0.1248, simple_loss=0.2066, pruned_loss=0.02146, over 1125255.42 frames. ], batch size: 54, lr: 3.31e-03, grad_scale: 8.0 +2023-03-21 13:50:28,105 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138678.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:50:29,646 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138681.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:50:32,153 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138686.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:50:34,079 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0339W0266-38957-0_sp0.9 from training. Duration: 12.5200625 +2023-03-21 13:50:38,217 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138697.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:50:52,679 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.1601, 2.6247, 1.9772, 2.9289, 2.7901, 3.1199, 2.7396, 2.6585], + device='cuda:1'), covar=tensor([0.2423, 0.1277, 0.4210, 0.1032, 0.0441, 0.0423, 0.0509, 0.0499], + device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0228, 0.0241, 0.0251, 0.0203, 0.0204, 0.0218, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:50:53,982 INFO [train.py:901] (1/2) Epoch 50, batch 350, loss[loss=0.1304, simple_loss=0.2108, pruned_loss=0.02495, over 7310.00 frames. ], tot_loss[loss=0.125, simple_loss=0.2067, pruned_loss=0.02162, over 1193113.63 frames. ], batch size: 59, lr: 3.31e-03, grad_scale: 8.0 +2023-03-21 13:51:01,762 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138742.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:51:04,310 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138747.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:51:08,202 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0249W0458-40379-0_sp0.9 from training. Duration: 12.348875 +2023-03-21 13:51:09,837 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138758.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:51:14,665 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.024e+02 1.762e+02 2.045e+02 2.495e+02 3.731e+02, threshold=4.091e+02, percent-clipped=0.0 +2023-03-21 13:51:19,715 INFO [train.py:901] (1/2) Epoch 50, batch 400, loss[loss=0.1356, simple_loss=0.216, pruned_loss=0.02766, over 7354.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.2076, pruned_loss=0.02193, over 1249099.66 frames. ], batch size: 51, lr: 3.31e-03, grad_scale: 8.0 +2023-03-21 13:51:21,874 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7936, 2.0031, 1.5816, 1.8033, 1.9631, 1.7353, 1.8486, 1.4236], + device='cuda:1'), covar=tensor([0.0216, 0.0344, 0.0576, 0.0301, 0.0244, 0.0172, 0.0261, 0.0286], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0042, 0.0040, 0.0042, 0.0039, 0.0039, 0.0041, 0.0051], + device='cuda:1'), out_proj_covar=tensor([4.8309e-05, 4.6331e-05, 4.5451e-05, 4.5938e-05, 4.3052e-05, 4.3219e-05, + 4.6382e-05, 5.5539e-05], device='cuda:1') +2023-03-21 13:51:22,163 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 +2023-03-21 13:51:26,332 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3377, 4.1444, 3.7579, 3.8586, 3.3412, 2.4394, 1.9701, 4.2651], + device='cuda:1'), covar=tensor([0.0049, 0.0079, 0.0107, 0.0088, 0.0155, 0.0557, 0.0690, 0.0058], + device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0095, 0.0119, 0.0102, 0.0136, 0.0137, 0.0133, 0.0109], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 13:51:34,877 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.5108, 2.4545, 2.4864, 2.4273, 2.8091, 2.5037, 2.4328, 2.1238], + device='cuda:1'), covar=tensor([0.0455, 0.0595, 0.0472, 0.0379, 0.0449, 0.0624, 0.0326, 0.0402], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0044, 0.0044, 0.0043, 0.0041, 0.0041, 0.0047, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 13:51:45,913 INFO [train.py:901] (1/2) Epoch 50, batch 450, loss[loss=0.1233, simple_loss=0.2117, pruned_loss=0.01742, over 7315.00 frames. ], tot_loss[loss=0.1258, simple_loss=0.2072, pruned_loss=0.02215, over 1292073.44 frames. ], batch size: 59, lr: 3.31e-03, grad_scale: 8.0 +2023-03-21 13:51:50,420 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0657W0266-27035-0_sp0.9 from training. Duration: 12.0923125 +2023-03-21 13:51:50,434 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0_sp0.9 from training. Duration: 13.955625 +2023-03-21 13:52:07,178 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.180e+02 1.662e+02 2.041e+02 2.387e+02 3.807e+02, threshold=4.081e+02, percent-clipped=0.0 +2023-03-21 13:52:11,756 INFO [train.py:901] (1/2) Epoch 50, batch 500, loss[loss=0.1619, simple_loss=0.2366, pruned_loss=0.04362, over 7264.00 frames. ], tot_loss[loss=0.126, simple_loss=0.2072, pruned_loss=0.02241, over 1323610.26 frames. ], batch size: 70, lr: 3.31e-03, grad_scale: 8.0 +2023-03-21 13:52:24,929 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0 from training. Duration: 13.031 +2023-03-21 13:52:26,420 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0269-85831-0_sp0.9 from training. Duration: 13.694375 +2023-03-21 13:52:26,926 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0421W0442-98380-0_sp0.9 from training. Duration: 13.187625 +2023-03-21 13:52:27,740 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 +2023-03-21 13:52:28,528 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138909.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:52:29,414 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0178W0358-85850-0_sp0.9 from training. Duration: 13.2424375 +2023-03-21 13:52:32,942 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0 from training. Duration: 14.53125 +2023-03-21 13:52:36,996 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138926.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:52:37,398 INFO [train.py:901] (1/2) Epoch 50, batch 550, loss[loss=0.127, simple_loss=0.2118, pruned_loss=0.0211, over 7324.00 frames. ], tot_loss[loss=0.125, simple_loss=0.2064, pruned_loss=0.02177, over 1349410.08 frames. ], batch size: 59, lr: 3.31e-03, grad_scale: 8.0 +2023-03-21 13:52:38,547 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2058, 2.2667, 2.3567, 2.1602, 2.5049, 2.2813, 2.1612, 1.7661], + device='cuda:1'), covar=tensor([0.0326, 0.0431, 0.0425, 0.0320, 0.0362, 0.0339, 0.0357, 0.0435], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0043, 0.0044, 0.0043, 0.0041, 0.0041, 0.0047, 0.0048], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 13:52:42,562 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8303, 3.0392, 3.7353, 3.7364, 3.8272, 3.8749, 3.8003, 3.7886], + device='cuda:1'), covar=tensor([0.0029, 0.0137, 0.0034, 0.0037, 0.0029, 0.0027, 0.0044, 0.0046], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0076, 0.0063, 0.0061, 0.0058, 0.0064, 0.0050, 0.0083], + device='cuda:1'), out_proj_covar=tensor([8.7735e-05, 1.4894e-04, 1.0850e-04, 1.0046e-04, 9.4542e-05, 1.0655e-04, + 9.0902e-05, 1.4899e-04], device='cuda:1') +2023-03-21 13:52:44,466 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0 from training. Duration: 12.062 +2023-03-21 13:52:52,560 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0723W0457-18562-0_sp0.9 from training. Duration: 12.26775 +2023-03-21 13:52:56,155 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0127W0333-18886-0_sp0.9 from training. Duration: 12.32125 +2023-03-21 13:52:58,589 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.346e+02 1.767e+02 2.083e+02 2.460e+02 6.205e+02, threshold=4.166e+02, percent-clipped=1.0 +2023-03-21 13:52:59,792 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138970.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 13:53:01,752 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=138974.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:53:03,244 INFO [train.py:901] (1/2) Epoch 50, batch 600, loss[loss=0.1462, simple_loss=0.2232, pruned_loss=0.0346, over 6616.00 frames. ], tot_loss[loss=0.1249, simple_loss=0.2061, pruned_loss=0.02182, over 1368529.93 frames. ], batch size: 107, lr: 3.31e-03, grad_scale: 8.0 +2023-03-21 13:53:03,270 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0227-102578-0_sp0.9 from training. Duration: 12.063375 +2023-03-21 13:53:03,849 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138978.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:53:15,981 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.8904, 4.3990, 4.2216, 4.8338, 4.6420, 4.7893, 4.3099, 4.4242], + device='cuda:1'), covar=tensor([0.0771, 0.2311, 0.2194, 0.0882, 0.0886, 0.1042, 0.0761, 0.1016], + device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0401, 0.0299, 0.0315, 0.0239, 0.0376, 0.0237, 0.0285], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 13:53:21,340 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0147-14735-0_sp0.9 from training. Duration: 13.0755 +2023-03-21 13:53:28,996 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139026.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:53:29,450 INFO [train.py:901] (1/2) Epoch 50, batch 650, loss[loss=0.1085, simple_loss=0.1812, pruned_loss=0.01788, over 7076.00 frames. ], tot_loss[loss=0.1252, simple_loss=0.2062, pruned_loss=0.02209, over 1384191.74 frames. ], batch size: 35, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 13:53:29,964 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0465-86257-0_sp0.9 from training. Duration: 12.8344375 +2023-03-21 13:53:34,565 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139037.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:53:37,050 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139042.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:53:43,233 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139053.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:53:47,825 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0704W0234-59743-0_sp0.9 from training. Duration: 12.6989375 +2023-03-21 13:53:50,871 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.276e+02 1.832e+02 2.103e+02 2.420e+02 4.796e+02, threshold=4.205e+02, percent-clipped=1.0 +2023-03-21 13:53:56,185 INFO [train.py:901] (1/2) Epoch 50, batch 700, loss[loss=0.1266, simple_loss=0.2019, pruned_loss=0.02562, over 7301.00 frames. ], tot_loss[loss=0.1256, simple_loss=0.2066, pruned_loss=0.02224, over 1396537.34 frames. ], batch size: 49, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 13:53:56,216 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0312-45768-0_sp0.9 from training. Duration: 12.422125 +2023-03-21 13:54:09,611 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7402, 2.3690, 2.9749, 2.8907, 3.0697, 2.7226, 2.4107, 2.9617], + device='cuda:1'), covar=tensor([0.1547, 0.1057, 0.0946, 0.0972, 0.0606, 0.1030, 0.1990, 0.0901], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0073, 0.0055, 0.0054, 0.0054, 0.0053, 0.0071, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 13:54:20,596 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0173-32335-0_sp0.9 from training. Duration: 13.2133125 +2023-03-21 13:54:21,044 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0247-36021-0_sp0.9 from training. Duration: 13.40225 +2023-03-21 13:54:21,535 INFO [train.py:901] (1/2) Epoch 50, batch 750, loss[loss=0.1117, simple_loss=0.1997, pruned_loss=0.01183, over 7343.00 frames. ], tot_loss[loss=0.1256, simple_loss=0.2067, pruned_loss=0.02231, over 1408168.58 frames. ], batch size: 63, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 13:54:28,343 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.0558, 2.6357, 3.2118, 3.0215, 3.2630, 2.9957, 2.7643, 3.2172], + device='cuda:1'), covar=tensor([0.1386, 0.0775, 0.0883, 0.1148, 0.0641, 0.0846, 0.1613, 0.0834], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0073, 0.0055, 0.0054, 0.0054, 0.0053, 0.0071, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 13:54:29,539 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.75 vs. limit=5.0 +2023-03-21 13:54:32,877 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0074, 4.3831, 4.3074, 4.9296, 4.7705, 4.8095, 4.3373, 4.4445], + device='cuda:1'), covar=tensor([0.0729, 0.2756, 0.2344, 0.0999, 0.0933, 0.1200, 0.0859, 0.1192], + device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0402, 0.0303, 0.0317, 0.0241, 0.0380, 0.0238, 0.0287], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], + device='cuda:1') +2023-03-21 13:54:34,856 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0064W0442-15905-0_sp0.9 from training. Duration: 12.0798125 +2023-03-21 13:54:38,900 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0062W0390-112800-0_sp0.9 from training. Duration: 12.335625 +2023-03-21 13:54:43,658 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.310e+02 1.826e+02 2.099e+02 2.412e+02 4.667e+02, threshold=4.199e+02, percent-clipped=1.0 +2023-03-21 13:54:45,226 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0 from training. Duration: 12.6340625 +2023-03-21 13:54:46,612 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp0.9 from training. Duration: 14.8566875 +2023-03-21 13:54:48,067 INFO [train.py:901] (1/2) Epoch 50, batch 800, loss[loss=0.1271, simple_loss=0.2118, pruned_loss=0.02119, over 6674.00 frames. ], tot_loss[loss=0.1262, simple_loss=0.2073, pruned_loss=0.02256, over 1416313.26 frames. ], batch size: 107, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 13:54:57,655 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0596W0136-102670-0_sp0.9 from training. Duration: 13.538875 +2023-03-21 13:55:14,432 INFO [train.py:901] (1/2) Epoch 50, batch 850, loss[loss=0.1036, simple_loss=0.181, pruned_loss=0.01313, over 7177.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.2067, pruned_loss=0.02252, over 1421992.83 frames. ], batch size: 39, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 13:55:16,899 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0_sp0.9 from training. Duration: 13.7823125 +2023-03-21 13:55:16,903 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0386-52709-0_sp0.9 from training. Duration: 12.77775 +2023-03-21 13:55:20,062 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139238.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:55:22,534 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0078W0423-41503-0_sp0.9 from training. Duration: 12.9845 +2023-03-21 13:55:26,080 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0462-112684-0_sp0.9 from training. Duration: 12.15225 +2023-03-21 13:55:34,313 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139265.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 13:55:35,726 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.240e+02 1.836e+02 2.170e+02 2.399e+02 3.753e+02, threshold=4.340e+02, percent-clipped=0.0 +2023-03-21 13:55:39,978 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139276.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:55:40,321 INFO [train.py:901] (1/2) Epoch 50, batch 900, loss[loss=0.1246, simple_loss=0.2062, pruned_loss=0.02155, over 7317.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2075, pruned_loss=0.02264, over 1428808.99 frames. ], batch size: 83, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 13:55:46,544 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139289.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:55:51,644 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139299.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:56:03,828 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0119W0370-86573-0_sp0.9 from training. Duration: 13.3566875 +2023-03-21 13:56:06,345 INFO [train.py:901] (1/2) Epoch 50, batch 950, loss[loss=0.1301, simple_loss=0.2114, pruned_loss=0.02441, over 7272.00 frames. ], tot_loss[loss=0.126, simple_loss=0.2071, pruned_loss=0.02242, over 1434113.07 frames. ], batch size: 52, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 13:56:11,568 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139337.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:56:11,603 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139337.0, num_to_drop=1, layers_to_drop={1} +2023-03-21 13:56:14,667 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139342.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:56:18,660 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139350.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:56:20,205 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139353.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:56:27,649 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.735e+02 1.986e+02 2.308e+02 3.681e+02, threshold=3.972e+02, percent-clipped=0.0 +2023-03-21 13:56:28,707 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0_sp0.9 from training. Duration: 14.10775 +2023-03-21 13:56:32,150 INFO [train.py:901] (1/2) Epoch 50, batch 1000, loss[loss=0.1358, simple_loss=0.2151, pruned_loss=0.02822, over 7257.00 frames. ], tot_loss[loss=0.1261, simple_loss=0.2071, pruned_loss=0.02253, over 1435659.75 frames. ], batch size: 55, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 13:56:36,240 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139385.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:56:38,759 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139390.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:56:45,301 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139401.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:56:49,819 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0027W0202-6154-0_sp0.9 from training. Duration: 12.862125 +2023-03-21 13:56:58,108 INFO [train.py:901] (1/2) Epoch 50, batch 1050, loss[loss=0.1239, simple_loss=0.2102, pruned_loss=0.01883, over 7281.00 frames. ], tot_loss[loss=0.1256, simple_loss=0.2068, pruned_loss=0.02218, over 1437451.08 frames. ], batch size: 66, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 13:57:11,472 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0480-79636-0_sp0.9 from training. Duration: 14.478875 +2023-03-21 13:57:14,423 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 +2023-03-21 13:57:15,528 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0600W0231-469-0_sp0.9 from training. Duration: 12.9744375 +2023-03-21 13:57:16,118 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.7706, 2.3741, 2.8805, 2.6984, 2.9679, 2.6905, 2.4080, 2.8233], + device='cuda:1'), covar=tensor([0.1338, 0.0835, 0.1320, 0.1334, 0.0755, 0.1258, 0.1725, 0.1310], + device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0072, 0.0054, 0.0054, 0.0053, 0.0053, 0.0071, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 13:57:19,436 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 1.721e+02 1.925e+02 2.210e+02 3.439e+02, threshold=3.850e+02, percent-clipped=0.0 +2023-03-21 13:57:23,956 INFO [train.py:901] (1/2) Epoch 50, batch 1100, loss[loss=0.1158, simple_loss=0.2062, pruned_loss=0.01267, over 7286.00 frames. ], tot_loss[loss=0.1252, simple_loss=0.2064, pruned_loss=0.022, over 1438455.20 frames. ], batch size: 68, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 13:57:45,148 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0007W0369-32499-0_sp0.9 from training. Duration: 12.25775 +2023-03-21 13:57:45,161 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0478-86278-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 13:57:45,313 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3064, 2.8155, 2.0574, 3.1938, 3.0267, 3.3467, 2.8043, 2.7474], + device='cuda:1'), covar=tensor([0.2337, 0.1128, 0.4210, 0.0806, 0.0403, 0.0343, 0.0465, 0.0566], + device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0230, 0.0242, 0.0251, 0.0204, 0.0204, 0.0218, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:57:47,831 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2459, 2.7278, 1.9660, 3.0385, 2.9166, 3.2174, 2.8209, 2.7087], + device='cuda:1'), covar=tensor([0.2467, 0.1169, 0.4142, 0.0706, 0.0424, 0.0338, 0.0509, 0.0534], + device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0230, 0.0242, 0.0251, 0.0204, 0.0204, 0.0218, 0.0226], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 13:57:50,644 INFO [train.py:901] (1/2) Epoch 50, batch 1150, loss[loss=0.1222, simple_loss=0.2091, pruned_loss=0.01764, over 7334.00 frames. ], tot_loss[loss=0.1251, simple_loss=0.2064, pruned_loss=0.02187, over 1438287.01 frames. ], batch size: 61, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 13:57:58,653 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0459-40844-0_sp0.9 from training. Duration: 13.11575 +2023-03-21 13:57:59,159 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0094W0460-59915-0_sp0.9 from training. Duration: 12.979125 +2023-03-21 13:58:09,851 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139565.0, num_to_drop=1, layers_to_drop={2} +2023-03-21 13:58:11,263 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.822e+02 2.028e+02 2.488e+02 3.805e+02, threshold=4.056e+02, percent-clipped=0.0 +2023-03-21 13:58:12,957 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139571.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:58:16,481 INFO [train.py:901] (1/2) Epoch 50, batch 1200, loss[loss=0.1057, simple_loss=0.1806, pruned_loss=0.01541, over 6947.00 frames. ], tot_loss[loss=0.125, simple_loss=0.2064, pruned_loss=0.02178, over 1440084.67 frames. ], batch size: 35, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 13:58:19,201 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139582.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:58:25,248 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139594.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:58:33,672 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0361W0207-97658-0_sp0.9 from training. Duration: 12.3268125 +2023-03-21 13:58:35,788 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139613.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:58:42,755 INFO [train.py:901] (1/2) Epoch 50, batch 1250, loss[loss=0.1331, simple_loss=0.2167, pruned_loss=0.02473, over 7320.00 frames. ], tot_loss[loss=0.1253, simple_loss=0.2065, pruned_loss=0.02205, over 1440585.54 frames. ], batch size: 75, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 13:58:45,423 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139632.0, num_to_drop=1, layers_to_drop={3} +2023-03-21 13:58:45,488 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139632.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:58:51,063 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139643.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:58:51,984 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139645.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 13:58:55,949 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0251-86481-0_sp0.9 from training. Duration: 13.1033125 +2023-03-21 13:59:01,058 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0039W0307-49457-0 from training. Duration: 12.4040625 +2023-03-21 13:59:02,072 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0121-45607-0_sp0.9 from training. Duration: 14.037875 +2023-03-21 13:59:03,970 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.064e+02 1.670e+02 1.886e+02 2.283e+02 6.089e+02, threshold=3.771e+02, percent-clipped=1.0 +2023-03-21 13:59:08,570 INFO [train.py:901] (1/2) Epoch 50, batch 1300, loss[loss=0.1408, simple_loss=0.2208, pruned_loss=0.03034, over 7265.00 frames. ], tot_loss[loss=0.1251, simple_loss=0.2067, pruned_loss=0.02175, over 1443850.24 frames. ], batch size: 64, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 13:59:24,800 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0013W0486-15004-0_sp0.9 from training. Duration: 12.0100625 +2023-03-21 13:59:27,673 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0703W0471-112404-0_sp0.9 from training. Duration: 12.43325 +2023-03-21 13:59:31,190 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0715W0442-101321-0 from training. Duration: 12.697 +2023-03-21 13:59:34,258 INFO [train.py:901] (1/2) Epoch 50, batch 1350, loss[loss=0.1152, simple_loss=0.1965, pruned_loss=0.01699, over 7317.00 frames. ], tot_loss[loss=0.125, simple_loss=0.2064, pruned_loss=0.02174, over 1444930.38 frames. ], batch size: 49, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 13:59:41,907 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0_sp0.9 from training. Duration: 13.930125 +2023-03-21 13:59:55,537 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.225e+02 1.698e+02 1.919e+02 2.403e+02 4.718e+02, threshold=3.837e+02, percent-clipped=1.0 +2023-03-21 13:59:56,181 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139769.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:00:00,123 INFO [train.py:901] (1/2) Epoch 50, batch 1400, loss[loss=0.1467, simple_loss=0.2241, pruned_loss=0.03464, over 7321.00 frames. ], tot_loss[loss=0.1249, simple_loss=0.2064, pruned_loss=0.02169, over 1443777.87 frames. ], batch size: 49, lr: 3.30e-03, grad_scale: 8.0 +2023-03-21 14:00:07,014 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139789.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:00:15,019 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139804.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:00:16,004 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139806.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:00:16,438 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0149-113139-0_sp0.9 from training. Duration: 12.431125 +2023-03-21 14:00:26,409 INFO [train.py:901] (1/2) Epoch 50, batch 1450, loss[loss=0.1299, simple_loss=0.2085, pruned_loss=0.02565, over 7277.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.2074, pruned_loss=0.02214, over 1447359.01 frames. ], batch size: 52, lr: 3.30e-03, grad_scale: 16.0 +2023-03-21 14:00:28,120 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139830.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:00:38,785 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:00:38,800 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:00:40,600 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0491-31830-0_sp0.9 from training. Duration: 13.241125 +2023-03-21 14:00:46,255 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139865.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:00:47,264 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139867.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:00:48,098 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.906e+02 2.192e+02 2.677e+02 5.370e+02, threshold=4.384e+02, percent-clipped=2.0 +2023-03-21 14:00:52,790 INFO [train.py:901] (1/2) Epoch 50, batch 1500, loss[loss=0.1397, simple_loss=0.2158, pruned_loss=0.03186, over 7339.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.2074, pruned_loss=0.02201, over 1446817.12 frames. ], batch size: 61, lr: 3.29e-03, grad_scale: 8.0 +2023-03-21 14:00:55,847 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp1.1 from training. Duration: 12.7053125 +2023-03-21 14:01:01,416 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139894.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:01:03,020 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9054, 3.3048, 2.1497, 3.2954, 2.5653, 2.8662, 1.4862, 2.3135], + device='cuda:1'), covar=tensor([0.0691, 0.1065, 0.3272, 0.0968, 0.0786, 0.0858, 0.4755, 0.2176], + device='cuda:1'), in_proj_covar=tensor([0.0271, 0.0252, 0.0269, 0.0262, 0.0261, 0.0258, 0.0222, 0.0249], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 14:01:10,140 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139911.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:01:11,145 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.4130, 4.8454, 4.8589, 4.8939, 4.7742, 4.4242, 4.9238, 4.7411], + device='cuda:1'), covar=tensor([0.0461, 0.0367, 0.0410, 0.0391, 0.0387, 0.0431, 0.0333, 0.0458], + device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0279, 0.0218, 0.0213, 0.0168, 0.0244, 0.0226, 0.0154], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 14:01:18,744 INFO [train.py:901] (1/2) Epoch 50, batch 1550, loss[loss=0.1211, simple_loss=0.2033, pruned_loss=0.01949, over 7291.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.2074, pruned_loss=0.022, over 1445432.41 frames. ], batch size: 68, lr: 3.29e-03, grad_scale: 8.0 +2023-03-21 14:01:18,815 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139927.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:01:19,817 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0 from training. Duration: 13.371 +2023-03-21 14:01:21,413 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139932.0, num_to_drop=1, layers_to_drop={0} +2023-03-21 14:01:24,346 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139938.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:01:26,369 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139942.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:01:27,918 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139945.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:01:40,572 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.234e+02 1.753e+02 2.129e+02 2.590e+02 7.618e+02, threshold=4.259e+02, percent-clipped=2.0 +2023-03-21 14:01:44,745 INFO [train.py:901] (1/2) Epoch 50, batch 1600, loss[loss=0.1287, simple_loss=0.2114, pruned_loss=0.02303, over 7262.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.2075, pruned_loss=0.02194, over 1447154.26 frames. ], batch size: 52, lr: 3.29e-03, grad_scale: 8.0 +2023-03-21 14:01:46,300 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139980.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:01:50,214 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0360W0244-86520-0_sp0.9 from training. Duration: 12.86675 +2023-03-21 14:01:50,758 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0434-40735-0 from training. Duration: 12.5600625 +2023-03-21 14:01:52,809 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=139993.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:01:54,719 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0_sp0.9 from training. Duration: 15.5286875 +2023-03-21 14:02:08,729 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0 from training. Duration: 13.074 +2023-03-21 14:02:11,764 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0162W0314-19617-0 from training. Duration: 12.1530625 +2023-03-21 14:02:14,262 INFO [train.py:901] (1/2) Epoch 50, batch 1650, loss[loss=0.1104, simple_loss=0.1921, pruned_loss=0.01431, over 7133.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2085, pruned_loss=0.0225, over 1444782.04 frames. ], batch size: 41, lr: 3.29e-03, grad_scale: 8.0 +2023-03-21 14:02:20,756 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0023W0395-77944-0_sp0.9 from training. Duration: 13.130125 +2023-03-21 14:02:32,808 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.8735, 2.5020, 3.1199, 2.8991, 3.0444, 2.8073, 2.5276, 2.7551], + device='cuda:1'), covar=tensor([0.1223, 0.0799, 0.0986, 0.1451, 0.0586, 0.0748, 0.1815, 0.1517], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0071, 0.0054, 0.0052, 0.0053, 0.0052, 0.0069, 0.0053], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:1') +2023-03-21 14:02:36,150 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.257e+02 1.886e+02 2.154e+02 2.450e+02 4.461e+02, threshold=4.308e+02, percent-clipped=1.0 +2023-03-21 14:02:37,729 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0713W0320-44227-0_sp0.9 from training. Duration: 12.868875 +2023-03-21 14:02:40,217 INFO [train.py:901] (1/2) Epoch 50, batch 1700, loss[loss=0.1215, simple_loss=0.2049, pruned_loss=0.01908, over 7277.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.2071, pruned_loss=0.0222, over 1440145.62 frames. ], batch size: 77, lr: 3.29e-03, grad_scale: 8.0 +2023-03-21 14:02:43,326 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0012W0478-21379-0_sp0.9 from training. Duration: 12.137875 +2023-03-21 14:02:44,437 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([5.0450, 5.5075, 5.5405, 5.5476, 5.2470, 5.0436, 5.6006, 5.4036], + device='cuda:1'), covar=tensor([0.0370, 0.0333, 0.0367, 0.0408, 0.0390, 0.0375, 0.0278, 0.0434], + device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0281, 0.0219, 0.0216, 0.0169, 0.0246, 0.0228, 0.0155], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 14:02:47,751 INFO [scaling.py:679] (1/2) Whitening: num_groups=1, num_channels=384, metric=3.37 vs. limit=5.0 +2023-03-21 14:02:53,570 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0_sp0.9 from training. Duration: 14.19225 +2023-03-21 14:02:54,232 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140103.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:02:56,215 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0619, 3.2994, 4.0519, 3.9310, 4.1889, 4.0514, 4.1228, 4.0004], + device='cuda:1'), covar=tensor([0.0029, 0.0126, 0.0030, 0.0041, 0.0024, 0.0032, 0.0038, 0.0048], + device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0076, 0.0064, 0.0062, 0.0059, 0.0064, 0.0051, 0.0085], + device='cuda:1'), out_proj_covar=tensor([8.8467e-05, 1.4957e-04, 1.1037e-04, 1.0167e-04, 9.5971e-05, 1.0799e-04, + 9.2652e-05, 1.5239e-04], device='cuda:1') +2023-03-21 14:02:56,529 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 +2023-03-21 14:02:57,780 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140110.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:03:02,407 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.0604, 2.5318, 2.0233, 2.7602, 2.7831, 2.6074, 2.6316, 2.4231], + device='cuda:1'), covar=tensor([0.2383, 0.1324, 0.4047, 0.0664, 0.0438, 0.0394, 0.0469, 0.0466], + device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0229, 0.0239, 0.0250, 0.0203, 0.0203, 0.0217, 0.0224], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 14:03:05,999 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140125.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:03:06,944 INFO [train.py:901] (1/2) Epoch 50, batch 1750, loss[loss=0.1265, simple_loss=0.2095, pruned_loss=0.02171, over 7284.00 frames. ], tot_loss[loss=0.1253, simple_loss=0.2067, pruned_loss=0.02194, over 1440086.66 frames. ], batch size: 68, lr: 3.29e-03, grad_scale: 8.0 +2023-03-21 14:03:12,156 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2836, 2.6534, 2.5907, 2.5888, 2.6661, 2.4923, 2.4148, 1.9883], + device='cuda:1'), covar=tensor([0.0694, 0.0489, 0.0676, 0.0205, 0.0668, 0.0523, 0.0443, 0.0538], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0043, 0.0044, 0.0043, 0.0041, 0.0041, 0.0048, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 14:03:16,125 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140145.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:03:18,009 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0025W0444-36635-0_sp0.9 from training. Duration: 13.15675 +2023-03-21 14:03:19,051 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0063W0488-49084-0_sp0.9 from training. Duration: 12.4045 +2023-03-21 14:03:22,762 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8598, 3.8558, 2.8886, 3.5141, 2.7668, 2.1026, 1.7716, 3.8191], + device='cuda:1'), covar=tensor([0.0071, 0.0078, 0.0227, 0.0106, 0.0259, 0.0705, 0.0831, 0.0088], + device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0097, 0.0122, 0.0104, 0.0140, 0.0139, 0.0136, 0.0111], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:1') +2023-03-21 14:03:23,719 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140160.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:03:24,734 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140162.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:03:25,801 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140164.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:03:28,147 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 1.791e+02 2.014e+02 2.461e+02 5.476e+02, threshold=4.028e+02, percent-clipped=1.0 +2023-03-21 14:03:29,348 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140171.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:03:32,258 INFO [train.py:901] (1/2) Epoch 50, batch 1800, loss[loss=0.1164, simple_loss=0.1926, pruned_loss=0.02009, over 7307.00 frames. ], tot_loss[loss=0.1252, simple_loss=0.2064, pruned_loss=0.022, over 1439992.18 frames. ], batch size: 49, lr: 3.29e-03, grad_scale: 8.0 +2023-03-21 14:03:41,470 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0242W0130-46331-0_sp0.9 from training. Duration: 13.184375 +2023-03-21 14:03:45,121 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8902, 1.7347, 2.1633, 2.2638, 2.0516, 2.2127, 1.8613, 2.3218], + device='cuda:1'), covar=tensor([0.2934, 0.3131, 0.2066, 0.2270, 0.1823, 0.1755, 0.1709, 0.2126], + device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0088, 0.0081, 0.0071, 0.0072, 0.0071, 0.0113, 0.0071], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 14:03:48,146 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140206.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:03:55,914 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0170W0469-46827-0_sp0.9 from training. Duration: 12.9201875 +2023-03-21 14:03:59,420 INFO [train.py:901] (1/2) Epoch 50, batch 1850, loss[loss=0.1031, simple_loss=0.1862, pruned_loss=0.009992, over 7182.00 frames. ], tot_loss[loss=0.1257, simple_loss=0.207, pruned_loss=0.02221, over 1440619.47 frames. ], batch size: 39, lr: 3.29e-03, grad_scale: 8.0 +2023-03-21 14:03:59,527 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140227.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:04:05,189 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140238.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:04:06,130 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0469-52699-0_sp0.9 from training. Duration: 12.1343125 +2023-03-21 14:04:15,861 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140259.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:04:21,421 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.209e+02 1.696e+02 1.990e+02 2.449e+02 3.903e+02, threshold=3.979e+02, percent-clipped=0.0 +2023-03-21 14:04:22,915 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0208-31683-0_sp0.9 from training. Duration: 12.273375 +2023-03-21 14:04:24,566 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140275.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:04:25,550 INFO [train.py:901] (1/2) Epoch 50, batch 1900, loss[loss=0.1161, simple_loss=0.2079, pruned_loss=0.01217, over 7305.00 frames. ], tot_loss[loss=0.1253, simple_loss=0.207, pruned_loss=0.02182, over 1442766.61 frames. ], batch size: 86, lr: 3.29e-03, grad_scale: 8.0 +2023-03-21 14:04:30,152 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140286.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:04:48,118 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140320.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:04:48,506 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0702W0289-75727-0_sp0.9 from training. Duration: 12.8045 +2023-03-21 14:04:51,574 INFO [train.py:901] (1/2) Epoch 50, batch 1950, loss[loss=0.1178, simple_loss=0.2005, pruned_loss=0.01755, over 7350.00 frames. ], tot_loss[loss=0.1248, simple_loss=0.2063, pruned_loss=0.02168, over 1443382.43 frames. ], batch size: 63, lr: 3.29e-03, grad_scale: 8.0 +2023-03-21 14:04:57,921 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-03-21 14:05:00,102 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0480-35915-0_sp0.9 from training. Duration: 12.51025 +2023-03-21 14:05:04,672 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0233W0325-47406-0_sp0.9 from training. Duration: 13.155375 +2023-03-21 14:05:05,176 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0708W0216-113208-0_sp1.1 from training. Duration: 12.1554375 +2023-03-21 14:05:13,389 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.395e+02 1.726e+02 1.971e+02 2.392e+02 3.404e+02, threshold=3.943e+02, percent-clipped=0.0 +2023-03-21 14:05:17,438 INFO [train.py:901] (1/2) Epoch 50, batch 2000, loss[loss=0.1271, simple_loss=0.2068, pruned_loss=0.02367, over 7242.00 frames. ], tot_loss[loss=0.1256, simple_loss=0.2073, pruned_loss=0.02196, over 1443922.66 frames. ], batch size: 55, lr: 3.29e-03, grad_scale: 8.0 +2023-03-21 14:05:22,441 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0212W0468-40835-0 from training. Duration: 13.9758125 +2023-03-21 14:05:22,542 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140387.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:05:33,996 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0472-41898-0_sp0.9 from training. Duration: 16.1458125 +2023-03-21 14:05:36,692 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8593, 2.1255, 1.7350, 1.8482, 1.9731, 1.8097, 2.0376, 1.7242], + device='cuda:1'), covar=tensor([0.0196, 0.0194, 0.0297, 0.0338, 0.0168, 0.0183, 0.0181, 0.0253], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0043, 0.0040, 0.0042, 0.0039, 0.0040, 0.0041, 0.0051], + device='cuda:1'), out_proj_covar=tensor([4.8736e-05, 4.7027e-05, 4.5119e-05, 4.6445e-05, 4.3205e-05, 4.4032e-05, + 4.6219e-05, 5.6206e-05], device='cuda:1') +2023-03-21 14:05:42,003 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0348W0263-105296-0_sp0.9 from training. Duration: 14.5266875 +2023-03-21 14:05:42,599 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140425.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:05:43,546 INFO [train.py:901] (1/2) Epoch 50, batch 2050, loss[loss=0.1162, simple_loss=0.1989, pruned_loss=0.01677, over 7322.00 frames. ], tot_loss[loss=0.1254, simple_loss=0.2065, pruned_loss=0.02209, over 1441417.52 frames. ], batch size: 49, lr: 3.29e-03, grad_scale: 8.0 +2023-03-21 14:05:52,771 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140445.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:05:54,815 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140448.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:06:00,290 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140459.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:06:00,852 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140460.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:06:01,872 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140462.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:06:03,820 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140466.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:06:05,184 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.175e+02 1.778e+02 2.115e+02 2.474e+02 4.579e+02, threshold=4.229e+02, percent-clipped=2.0 +2023-03-21 14:06:07,202 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140473.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:06:09,168 INFO [train.py:901] (1/2) Epoch 50, batch 2100, loss[loss=0.1411, simple_loss=0.2198, pruned_loss=0.03125, over 7264.00 frames. ], tot_loss[loss=0.1258, simple_loss=0.2069, pruned_loss=0.02236, over 1441493.13 frames. ], batch size: 64, lr: 3.29e-03, grad_scale: 8.0 +2023-03-21 14:06:14,934 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.6282, 2.8846, 2.5111, 2.8040, 2.7774, 2.5507, 2.7179, 2.6818], + device='cuda:1'), covar=tensor([0.0634, 0.0529, 0.1226, 0.0853, 0.0831, 0.0701, 0.0966, 0.0708], + device='cuda:1'), in_proj_covar=tensor([0.0063, 0.0064, 0.0072, 0.0064, 0.0060, 0.0068, 0.0061, 0.0057], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 14:06:15,924 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140489.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:06:16,803 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0190W0196-79669-0_sp0.9 from training. Duration: 12.3110625 +2023-03-21 14:06:17,872 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140493.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:06:19,356 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0364-52727-0 from training. Duration: 12.537125 +2023-03-21 14:06:24,455 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140506.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:06:25,390 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140508.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:06:26,025 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140509.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:06:26,426 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140510.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:06:34,791 INFO [train.py:901] (1/2) Epoch 50, batch 2150, loss[loss=0.1337, simple_loss=0.2209, pruned_loss=0.02323, over 7158.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.2072, pruned_loss=0.02231, over 1441867.10 frames. ], batch size: 98, lr: 3.29e-03, grad_scale: 8.0 +2023-03-21 14:06:47,334 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140550.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:06:49,233 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140554.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:06:50,322 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.8454, 1.7497, 2.2668, 2.3912, 2.1718, 2.2955, 2.0142, 2.3464], + device='cuda:1'), covar=tensor([0.2999, 0.4035, 0.1560, 0.1484, 0.1561, 0.1570, 0.2019, 0.3813], + device='cuda:1'), in_proj_covar=tensor([0.0089, 0.0088, 0.0082, 0.0072, 0.0072, 0.0071, 0.0114, 0.0073], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 14:06:57,443 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.260e+02 1.798e+02 2.077e+02 2.497e+02 4.046e+02, threshold=4.154e+02, percent-clipped=0.0 +2023-03-21 14:06:58,157 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140570.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:07:01,574 INFO [train.py:901] (1/2) Epoch 50, batch 2200, loss[loss=0.1275, simple_loss=0.2129, pruned_loss=0.02103, over 7267.00 frames. ], tot_loss[loss=0.1262, simple_loss=0.2075, pruned_loss=0.02242, over 1443109.00 frames. ], batch size: 70, lr: 3.29e-03, grad_scale: 8.0 +2023-03-21 14:07:05,148 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0286-35767-0_sp0.9 from training. Duration: 13.03575 +2023-03-21 14:07:10,859 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.1817, 3.4518, 2.3548, 3.6887, 2.8999, 3.2904, 1.6460, 2.5273], + device='cuda:1'), covar=tensor([0.0673, 0.1072, 0.3291, 0.0780, 0.0697, 0.0845, 0.4705, 0.2175], + device='cuda:1'), in_proj_covar=tensor([0.0272, 0.0254, 0.0271, 0.0263, 0.0263, 0.0260, 0.0223, 0.0250], + device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 14:07:17,265 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6857, 1.8773, 1.7142, 1.7517, 1.7817, 1.7717, 1.7646, 1.5372], + device='cuda:1'), covar=tensor([0.0183, 0.0222, 0.0404, 0.0243, 0.0147, 0.0198, 0.0182, 0.0240], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0043, 0.0040, 0.0042, 0.0039, 0.0040, 0.0041, 0.0051], + device='cuda:1'), out_proj_covar=tensor([4.8462e-05, 4.6996e-05, 4.5060e-05, 4.6341e-05, 4.3116e-05, 4.3934e-05, + 4.6227e-05, 5.5922e-05], device='cuda:1') +2023-03-21 14:07:21,243 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140615.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:07:27,034 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.7044, 1.6415, 1.7673, 1.9961, 1.8164, 2.0976, 1.5617, 1.9781], + device='cuda:1'), covar=tensor([0.1949, 0.3595, 0.1281, 0.1072, 0.1272, 0.0944, 0.2617, 0.2245], + device='cuda:1'), in_proj_covar=tensor([0.0089, 0.0088, 0.0082, 0.0072, 0.0072, 0.0071, 0.0114, 0.0073], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 14:07:27,881 INFO [train.py:901] (1/2) Epoch 50, batch 2250, loss[loss=0.1455, simple_loss=0.2222, pruned_loss=0.0344, over 7316.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2082, pruned_loss=0.02264, over 1444074.11 frames. ], batch size: 54, lr: 3.29e-03, grad_scale: 4.0 +2023-03-21 14:07:39,398 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0169W0400-1289-0_sp0.9 from training. Duration: 13.5110625 +2023-03-21 14:07:39,878 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0101W0466-14509-0_sp0.9 from training. Duration: 12.485625 +2023-03-21 14:07:50,016 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.217e+02 1.844e+02 2.128e+02 2.525e+02 5.922e+02, threshold=4.256e+02, percent-clipped=1.0 +2023-03-21 14:07:52,671 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0711W0333-38122-0_sp0.9 from training. Duration: 13.312125 +2023-03-21 14:07:53,660 INFO [train.py:901] (1/2) Epoch 50, batch 2300, loss[loss=0.1296, simple_loss=0.2153, pruned_loss=0.02197, over 7309.00 frames. ], tot_loss[loss=0.1268, simple_loss=0.2083, pruned_loss=0.02266, over 1442616.99 frames. ], batch size: 80, lr: 3.29e-03, grad_scale: 4.0 +2023-03-21 14:08:17,494 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140723.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:08:19,362 INFO [train.py:901] (1/2) Epoch 50, batch 2350, loss[loss=0.1273, simple_loss=0.2076, pruned_loss=0.0235, over 7313.00 frames. ], tot_loss[loss=0.1267, simple_loss=0.2079, pruned_loss=0.02274, over 1441441.75 frames. ], batch size: 75, lr: 3.28e-03, grad_scale: 4.0 +2023-03-21 14:08:28,110 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140743.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:08:36,198 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140759.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:08:39,749 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140766.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:08:40,176 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0014W0402-35914-0_sp0.9 from training. Duration: 13.08 +2023-03-21 14:08:41,683 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.344e+02 1.750e+02 2.041e+02 2.523e+02 5.512e+02, threshold=4.082e+02, percent-clipped=1.0 +2023-03-21 14:08:45,233 INFO [train.py:901] (1/2) Epoch 50, batch 2400, loss[loss=0.1327, simple_loss=0.2067, pruned_loss=0.02939, over 7297.00 frames. ], tot_loss[loss=0.1259, simple_loss=0.2069, pruned_loss=0.02249, over 1440972.11 frames. ], batch size: 49, lr: 3.28e-03, grad_scale: 8.0 +2023-03-21 14:08:46,290 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0337W0459-31834-0_sp0.9 from training. Duration: 13.051 +2023-03-21 14:08:48,954 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140784.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:08:58,359 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0215W0439-96772-0_sp0.9 from training. Duration: 12.086625 +2023-03-21 14:09:01,371 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0016W0371-52743-0_sp0.9 from training. Duration: 12.3335 +2023-03-21 14:09:01,408 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140807.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:09:02,493 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140809.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:09:05,003 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140814.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:09:08,635 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.2156, 2.3625, 2.3842, 2.3044, 2.6340, 2.4704, 2.3236, 1.9271], + device='cuda:1'), covar=tensor([0.0522, 0.0499, 0.0531, 0.0271, 0.0566, 0.0631, 0.0325, 0.0332], + device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0043, 0.0043, 0.0043, 0.0041, 0.0041, 0.0047, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 14:09:12,232 INFO [train.py:901] (1/2) Epoch 50, batch 2450, loss[loss=0.1159, simple_loss=0.2031, pruned_loss=0.0144, over 7323.00 frames. ], tot_loss[loss=0.1266, simple_loss=0.2074, pruned_loss=0.02293, over 1443169.90 frames. ], batch size: 59, lr: 3.28e-03, grad_scale: 8.0 +2023-03-21 14:09:21,032 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.0208, 4.1876, 3.9314, 4.1645, 3.7162, 4.0850, 4.4066, 4.4267], + device='cuda:1'), covar=tensor([0.0191, 0.0134, 0.0199, 0.0147, 0.0292, 0.0258, 0.0190, 0.0156], + device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0128, 0.0123, 0.0126, 0.0115, 0.0103, 0.0100, 0.0103], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 14:09:21,493 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140845.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:09:22,225 INFO [scaling.py:679] (1/2) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-03-21 14:09:28,611 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0521W0124-45606-0 from training. Duration: 12.773 +2023-03-21 14:09:31,784 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140865.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:09:34,208 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.235e+02 1.733e+02 1.983e+02 2.294e+02 8.300e+02, threshold=3.966e+02, percent-clipped=1.0 +2023-03-21 14:09:34,388 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140870.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:09:37,784 INFO [train.py:901] (1/2) Epoch 50, batch 2500, loss[loss=0.1134, simple_loss=0.1816, pruned_loss=0.02261, over 6991.00 frames. ], tot_loss[loss=0.1268, simple_loss=0.2076, pruned_loss=0.02297, over 1443781.98 frames. ], batch size: 35, lr: 3.28e-03, grad_scale: 8.0 +2023-03-21 14:09:50,524 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.3565, 2.8938, 2.2273, 3.0116, 3.2611, 3.3229, 3.0312, 2.8255], + device='cuda:1'), covar=tensor([0.2620, 0.1253, 0.4369, 0.0898, 0.0480, 0.0390, 0.0560, 0.0584], + device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0231, 0.0241, 0.0252, 0.0206, 0.0205, 0.0219, 0.0225], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 14:09:54,466 WARNING [train.py:1061] (1/2) Exclude cut with ID BAC009S0658W0494-41924-0_sp0.9 from training. Duration: 12.070375 +2023-03-21 14:09:57,587 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140915.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:09:58,637 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140917.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:10:04,307 INFO [train.py:901] (1/2) Epoch 50, batch 2550, loss[loss=0.1286, simple_loss=0.2153, pruned_loss=0.02092, over 7332.00 frames. ], tot_loss[loss=0.127, simple_loss=0.2082, pruned_loss=0.0229, over 1444678.16 frames. ], batch size: 59, lr: 3.28e-03, grad_scale: 8.0 +2023-03-21 14:10:14,626 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140947.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:10:19,736 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.5589, 1.6539, 1.4397, 1.5340, 1.5834, 1.5785, 1.4349, 1.4038], + device='cuda:1'), covar=tensor([0.0184, 0.0159, 0.0209, 0.0167, 0.0148, 0.0150, 0.0234, 0.0198], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0042, 0.0040, 0.0042, 0.0039, 0.0039, 0.0041, 0.0051], + device='cuda:1'), out_proj_covar=tensor([4.8161e-05, 4.6853e-05, 4.5081e-05, 4.5927e-05, 4.2842e-05, 4.3376e-05, + 4.6222e-05, 5.5412e-05], device='cuda:1') +2023-03-21 14:10:22,711 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=140963.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:10:26,237 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.104e+02 1.777e+02 2.057e+02 2.354e+02 4.545e+02, threshold=4.114e+02, percent-clipped=1.0 +2023-03-21 14:10:27,900 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.3729, 4.8869, 4.9662, 4.8943, 4.8086, 4.3954, 4.9817, 4.8078], + device='cuda:1'), covar=tensor([0.0445, 0.0385, 0.0416, 0.0459, 0.0365, 0.0445, 0.0370, 0.0416], + device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0280, 0.0218, 0.0214, 0.0167, 0.0245, 0.0225, 0.0153], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 14:10:30,291 INFO [train.py:901] (1/2) Epoch 50, batch 2600, loss[loss=0.127, simple_loss=0.2038, pruned_loss=0.02512, over 7294.00 frames. ], tot_loss[loss=0.1272, simple_loss=0.2084, pruned_loss=0.02297, over 1447130.13 frames. ], batch size: 86, lr: 3.28e-03, grad_scale: 8.0 +2023-03-21 14:10:30,984 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140978.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:10:40,929 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.4063, 2.6510, 2.7505, 2.3542, 2.7193, 2.5728, 2.3691, 1.9831], + device='cuda:1'), covar=tensor([0.0379, 0.0383, 0.0408, 0.0402, 0.0563, 0.0578, 0.0369, 0.0375], + device='cuda:1'), in_proj_covar=tensor([0.0045, 0.0043, 0.0043, 0.0043, 0.0041, 0.0041, 0.0047, 0.0047], + device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:1') +2023-03-21 14:10:46,202 INFO [zipformer.py:625] (1/2) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141008.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:10:46,724 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([1.6012, 1.7387, 1.5677, 1.6360, 1.7064, 1.7002, 1.5795, 1.4703], + device='cuda:1'), covar=tensor([0.0208, 0.0192, 0.0215, 0.0201, 0.0142, 0.0139, 0.0179, 0.0190], + device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0042, 0.0040, 0.0042, 0.0039, 0.0039, 0.0041, 0.0051], + device='cuda:1'), out_proj_covar=tensor([4.8171e-05, 4.6937e-05, 4.5197e-05, 4.5980e-05, 4.2902e-05, 4.3307e-05, + 4.6323e-05, 5.5458e-05], device='cuda:1') +2023-03-21 14:10:55,560 INFO [train.py:901] (1/2) Epoch 50, batch 2650, loss[loss=0.1277, simple_loss=0.2132, pruned_loss=0.02106, over 7294.00 frames. ], tot_loss[loss=0.1266, simple_loss=0.208, pruned_loss=0.02261, over 1444727.08 frames. ], batch size: 68, lr: 3.28e-03, grad_scale: 8.0 +2023-03-21 14:10:59,146 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([2.9820, 2.5550, 2.9541, 2.9968, 2.7545, 2.7957, 3.0666, 2.2599], + device='cuda:1'), covar=tensor([0.0510, 0.0630, 0.1056, 0.0728, 0.0742, 0.1026, 0.0640, 0.3166], + device='cuda:1'), in_proj_covar=tensor([0.0330, 0.0339, 0.0270, 0.0348, 0.0280, 0.0285, 0.0344, 0.0235], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:1') +2023-03-21 14:11:03,580 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141043.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:11:12,475 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.2777, 3.4735, 3.1192, 4.1136, 2.5484, 4.2012, 1.9834, 3.7156], + device='cuda:1'), covar=tensor([0.0171, 0.1007, 0.1609, 0.0250, 0.3928, 0.0302, 0.1275, 0.0476], + device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0242, 0.0255, 0.0215, 0.0247, 0.0220, 0.0216, 0.0230], + device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:1') +2023-03-21 14:11:16,709 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.285e+02 1.781e+02 2.065e+02 2.401e+02 3.842e+02, threshold=4.131e+02, percent-clipped=0.0 +2023-03-21 14:11:20,197 INFO [train.py:901] (1/2) Epoch 50, batch 2700, loss[loss=0.1331, simple_loss=0.2115, pruned_loss=0.02739, over 7293.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2077, pruned_loss=0.02256, over 1442300.90 frames. ], batch size: 49, lr: 3.28e-03, grad_scale: 8.0 +2023-03-21 14:11:21,242 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141079.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:11:27,662 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=141091.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:11:27,722 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([3.8618, 2.9507, 3.8332, 3.7803, 3.9929, 3.8492, 3.7548, 3.7417], + device='cuda:1'), covar=tensor([0.0027, 0.0142, 0.0030, 0.0033, 0.0026, 0.0030, 0.0061, 0.0054], + device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0075, 0.0063, 0.0061, 0.0059, 0.0064, 0.0050, 0.0085], + device='cuda:1'), out_proj_covar=tensor([8.6535e-05, 1.4771e-04, 1.0887e-04, 1.0052e-04, 9.5475e-05, 1.0768e-04, + 9.1373e-05, 1.5250e-04], device='cuda:1') +2023-03-21 14:11:38,470 INFO [zipformer.py:1455] (1/2) attn_weights_entropy = tensor([4.5327, 4.1104, 4.1011, 4.2641, 4.2737, 4.1212, 4.3884, 3.9058], + device='cuda:1'), covar=tensor([0.0146, 0.0158, 0.0147, 0.0151, 0.0393, 0.0143, 0.0151, 0.0169], + device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0107, 0.0109, 0.0093, 0.0181, 0.0113, 0.0109, 0.0119], + device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002], + device='cuda:1') +2023-03-21 14:11:45,199 INFO [train.py:901] (1/2) Epoch 50, batch 2750, loss[loss=0.1267, simple_loss=0.2061, pruned_loss=0.02368, over 7314.00 frames. ], tot_loss[loss=0.1264, simple_loss=0.2077, pruned_loss=0.02253, over 1441704.46 frames. ], batch size: 59, lr: 3.28e-03, grad_scale: 8.0 +2023-03-21 14:11:53,275 INFO [zipformer.py:625] (1/2) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141143.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:11:54,257 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141145.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:12:03,915 INFO [zipformer.py:625] (1/2) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:12:03,941 INFO [zipformer.py:625] (1/2) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:12:06,328 INFO [optim.py:369] (1/2) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 1.653e+02 1.885e+02 2.359e+02 4.895e+02, threshold=3.769e+02, percent-clipped=3.0 +2023-03-21 14:12:09,768 INFO [train.py:901] (1/2) Epoch 50, batch 2800, loss[loss=0.1382, simple_loss=0.2076, pruned_loss=0.03435, over 7263.00 frames. ], tot_loss[loss=0.127, simple_loss=0.2086, pruned_loss=0.02269, over 1441772.82 frames. ], batch size: 47, lr: 3.28e-03, grad_scale: 8.0 +2023-03-21 14:12:17,602 INFO [zipformer.py:625] (1/2) warmup_begin=666.7, warmup_end=1333.3, batch_count=141193.0, num_to_drop=0, layers_to_drop=set() +2023-03-21 14:12:21,741 INFO [train.py:1170] (1/2) Done!